var bibbase_data = {"data":"\"Loading..\"\n\n
\n\n \n\n \n\n \n \n\n \n\n \n \n\n \n\n\n
\n generated by\n \n \"bibbase.org\"\n\n \n
\n\n
\n\n \n\n\n
\n \n\n Excellent! Next, you can embed this page using\n one of several\n options.\n
\n\n
\n \n\n

To the site owner:

\n\n

Action required! Mendeley is changing its\n API. In order to keep using Mendeley with BibBase past April\n 14th, you need to:\n

    \n
  1. renew the authorization for BibBase on Mendeley, and
  2. \n
  3. update the BibBase URL\n in your page the same way you did when you initially set up\n this page.
  4. \n

\n\n

\n \n \n Fix it now\n

\n
\n\n
\n\n\n
\n \n \n
\n
\n  \n 2018\n \n \n (6)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n End-to-End Joint Semantic Segmentation of Actors and Actions in Video.\n \n\n\n \n Ji, J.; Buch, S.; Niebles, J.; and Soto, A.\n\n\n \n\n\n\n In ECCV, 2018. \n \n\n\n\n
\n\n\n\n \n \n \"End-to-EndPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 1 download\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Jingwei:EtAl:2018,\n  Author = {J. Ji and S. Buch and JC. Niebles and A. Soto},\n  Title = {End-to-End Joint Semantic Segmentation of Actors and Actions in Video},\n  booktitle = {{ECCV}},\n  year = {2018},\n  abstract = {Traditional video understanding tasks include human action recognition and actor-object semantic segmentation. However, the joint task of providing semantic segmentation for different actor classes simultaneously with their action class remains a challenging but necessary task for many applications. In this work, we propose a new end-to-end architecture for tackling this joint task in videos. Our model effectively leverages multiple input modalities, contextual information, and joint multitask learning in the video to directly output semantic segmentations in a single unified framework. We train and benchmark our model on the large-scale Actor-Action Dataset (A2D) for joint actor-action semantic segmentation, and demonstrate state-of-the-art performance for both segmentation and detection. We also perform experiments verifying our joint approach improves performance for zero-shot understanding, indicating generalizability of our jointly learned feature space.},\nurl={http://svl.stanford.edu/assets/papers/ji2018eccv.pdf},\n}\n\n\n
\n
\n\n\n
\n Traditional video understanding tasks include human action recognition and actor-object semantic segmentation. However, the joint task of providing semantic segmentation for different actor classes simultaneously with their action class remains a challenging but necessary task for many applications. In this work, we propose a new end-to-end architecture for tackling this joint task in videos. Our model effectively leverages multiple input modalities, contextual information, and joint multitask learning in the video to directly output semantic segmentations in a single unified framework. We train and benchmark our model on the large-scale Actor-Action Dataset (A2D) for joint actor-action semantic segmentation, and demonstrate state-of-the-art performance for both segmentation and detection. We also perform experiments verifying our joint approach improves performance for zero-shot understanding, indicating generalizability of our jointly learned feature space.\n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Translating Navigation Instructions in Natural Language to a High-Level Plan for Behavioral Robot Navigation.\n \n\n\n \n Zang, X.; Pokle, A.; Chen, K.; Vazquez, M.; Niebles, J.; Soto, A.; and Savaresse, S.\n\n\n \n\n\n\n In EMNLP, 2018. \n \n\n\n\n
\n\n\n\n \n \n \"TranslatingPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Xiaoxue:EtAl:2018,\n  Author = {X. Zang and Ashwini Pokle and Kevin Chen and M. Vazquez and JC. Niebles and A. Soto and S. Savaresse},\n  Title = {Translating Navigation Instructions in Natural Language to a High-Level Plan for Behavioral Robot Navigation},\n  booktitle = {EMNLP},\n  year = {2018},\n  abstract = {We propose an end-to-end deep learning model for translating free-form natural language instructions to a high-level plan for behavioral robot navigation. We use attention models to connect information from both the user instructions and a topological representation of the environment. We evaluate our model's performance on a new dataset containing 10,050 pairs of navigation instructions. Our model significantly outperforms baseline approaches. Furthermore, our results suggest that it is possible to leverage the environment map as a relevant knowledge base to facilitate the translation of free-form navigational instruction.},\nurl={https://arxiv.org/abs/1810.00663},\n}\n\n\n
\n
\n\n\n
\n We propose an end-to-end deep learning model for translating free-form natural language instructions to a high-level plan for behavioral robot navigation. We use attention models to connect information from both the user instructions and a topological representation of the environment. We evaluate our model's performance on a new dataset containing 10,050 pairs of navigation instructions. Our model significantly outperforms baseline approaches. Furthermore, our results suggest that it is possible to leverage the environment map as a relevant knowledge base to facilitate the translation of free-form navigational instruction.\n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Interpretable Visual Question Answering by Visual Grounding from Automatic Attention Annotations.\n \n\n\n \n Zhang, B.; Niebles, J.; and Soto, A.\n\n\n \n\n\n\n In WACV, 2018. \n \n\n\n\n
\n\n\n\n \n \n \"InterpretablePaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Ben:EtAl:2018,\n  Author = {B. Zhang and JC. Niebles and A. Soto},\n  Title = {Interpretable Visual Question Answering by Visual Grounding from Automatic Attention Annotations},\n  booktitle = {WACV},\n  year = {2018},\n  abstract = {A key aspect of VQA models that are interpretable is their ability to ground their answers to relevant regions in the image. Current approaches with this capability rely on supervised learning and human annotated groundings to train attention mechanisms inside the VQA architecture. Unfortunately, obtaining human annotations specific for visual grounding is difficult and expensive. In this work, we demonstrate that we can effectively train a VQA architecture with grounding supervision that can be automatically obtained from available region descriptions and object annotations. We also show that our model trained with this mined supervision generates visual groundings that achieve higher correlation to manually-annotated groundings than alternative approaches, even in the case of state-of-the-art algorithms that are directly trained with human grounding annotations.},\nurl={https://arxiv.org/abs/1808.00265},\n}\n\n\n
\n
\n\n\n
\n A key aspect of VQA models that are interpretable is their ability to ground their answers to relevant regions in the image. Current approaches with this capability rely on supervised learning and human annotated groundings to train attention mechanisms inside the VQA architecture. Unfortunately, obtaining human annotations specific for visual grounding is difficult and expensive. In this work, we demonstrate that we can effectively train a VQA architecture with grounding supervision that can be automatically obtained from available region descriptions and object annotations. We also show that our model trained with this mined supervision generates visual groundings that achieve higher correlation to manually-annotated groundings than alternative approaches, even in the case of state-of-the-art algorithms that are directly trained with human grounding annotations.\n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Behavioral Indoor Navigation With Natural Language Directions.\n \n\n\n \n Zang, X.; Vázquez, M.; Niebles, J.; Soto, A.; and Savarese, S.\n\n\n \n\n\n\n In HRI, 2018. \n \n\n\n\n
\n\n\n\n \n \n \"BehavioralPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Zang:EtAl:2018,\n  Author = {X. Zang and M. Vázquez and JC. Niebles and A. Soto and S. Savarese},\n  Title = {Behavioral Indoor Navigation With Natural Language Directions},\n  booktitle = {{HRI}},\n  year = {2018},\n  abstract = {We describe a behavioral navigation approach that leverages the\nrich semantic structure of human environments to enable robots to\nnavigate without an explicit geometric representation of the world.\nBased on this approach, we then present our efforts to allow robots\nto follow navigation instructions in natural language. With our\nproof-of-concept implementation, we were able to translate natural\nlanguage navigation commands into a sequence of behaviors that\ncould then be executed by a robot to reach a desired goal.},\nurl={http://www.marynel.net/static/pdfs/zang-HRI18.pdf},\n}\n\n\n
\n
\n\n\n
\n We describe a behavioral navigation approach that leverages the rich semantic structure of human environments to enable robots to navigate without an explicit geometric representation of the world. Based on this approach, we then present our efforts to allow robots to follow navigation instructions in natural language. With our proof-of-concept implementation, we were able to translate natural language navigation commands into a sequence of behaviors that could then be executed by a robot to reach a desired goal.\n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n A Deep Learning Based Behavioral Approach to Indoor Autonomous Navigation.\n \n\n\n \n Sepulveda, G.; Niebles, J.; and Soto, A.\n\n\n \n\n\n\n In ICRA, 2018. \n \n\n\n\n
\n\n\n\n \n \n \"APaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Sepulveda:EtAl:2018,\n  Author = {G. Sepulveda and JC. Niebles and A. Soto},\n  Title = {A Deep Learning Based Behavioral Approach to Indoor Autonomous Navigation},\n  booktitle = {{ICRA}},\n  year = {2018},\n  abstract = {We present a semantically rich graph representa-\ntion for indoor robotic navigation. Our graph representation\nencodes: semantic locations such as offices or corridors as\nnodes, and navigational behaviors such as enter office or cross\na corridor as edges. In particular, our navigational behaviors\noperate directly from visual inputs to produce motor controls\nand are implemented with deep learning architectures. This\nenables the robot to avoid explicit computation of its precise\nlocation or the geometry of the environment, and enables\nnavigation at a higher level of semantic abstraction. We evaluate\nthe effectiveness of our representation by simulating navigation\ntasks in a large number of virtual environments. Our results\nshow that using a simple sets of perceptual and navigational\nbehaviors, the proposed approach can successfully guide the\nway of the robot as it completes navigational missions such\nas going to a specific office. Furthermore, our implementation\nshows to be effective to control the selection and switching of\nbehaviors.\n},\nurl={https://arxiv.org/pdf/1803.04119v1.pdf},\n}\n\n\n
\n
\n\n\n
\n We present a semantically rich graph representa- tion for indoor robotic navigation. Our graph representation encodes: semantic locations such as offices or corridors as nodes, and navigational behaviors such as enter office or cross a corridor as edges. In particular, our navigational behaviors operate directly from visual inputs to produce motor controls and are implemented with deep learning architectures. This enables the robot to avoid explicit computation of its precise location or the geometry of the environment, and enables navigation at a higher level of semantic abstraction. We evaluate the effectiveness of our representation by simulating navigation tasks in a large number of virtual environments. Our results show that using a simple sets of perceptual and navigational behaviors, the proposed approach can successfully guide the way of the robot as it completes navigational missions such as going to a specific office. Furthermore, our implementation shows to be effective to control the selection and switching of behaviors. \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Content-based artwork recommendation: integrating painting metadata with neural and manually-engineered visual features.\n \n\n\n \n Messina, P.; Dominguez, V.; Parra, D.; Trattner, C.; and Soto, A.\n\n\n \n\n\n\n User Modeling and User-Adapted Interaction. 2018.\n \n\n\n\n
\n\n\n\n \n \n \"Content-basedPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 3 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Messina:EtAl:2017,\n  Author = {Pablo Messina and Vicente Dominguez and Denis Parra and Christoph Trattner and Alvaro Soto},\n  Title = {Content-based artwork recommendation: integrating painting metadata with neural and manually-engineered visual features},\n  Journal = {User Modeling and User-Adapted Interaction},\n  Year = {2018},\n  abstract = {Recommender Systems help us deal with information overload by suggesting relevant items based on our personal preferences. Although there is a large body of research in areas such as movies or music, artwork recommendation has received comparatively little attention, despite the continuous growth of the artwork market. Most previous research has relied on ratings and metadata, and a few recent works have exploited visual features extracted with deep neural networks (DNN) to recommend digital art. In this work, we contribute to the area of content-based artwork recommendation of physical paintings by studying the impact of the aforementioned features (artwork metadata, neural visual features), as well as manually-engineered visual features, such as naturalness, brightness and contrast. We implement and evaluate our method using transactional data from UGallery.com, an online artwork store. Our results show that artwork recommendations based on a hybrid combination of artist preference, curated attributes, deep neural visual features and manually-engineered visual features produce the best performance. Moreover, we discuss the trade-off between automatically obtained DNN features and manually-engineered visual features for the purpose of explainability, as well as the impact of user profile size on predictions. Our research informs the development of next-generation content-based artwork recommenders which rely on different types of data, from text to multimedia.},\nurl = {https://link.springer.com/article/10.1007/s11257-018-9206-9}\n}\n\n\n%***********2017***************%\n
\n
\n\n\n
\n Recommender Systems help us deal with information overload by suggesting relevant items based on our personal preferences. Although there is a large body of research in areas such as movies or music, artwork recommendation has received comparatively little attention, despite the continuous growth of the artwork market. Most previous research has relied on ratings and metadata, and a few recent works have exploited visual features extracted with deep neural networks (DNN) to recommend digital art. In this work, we contribute to the area of content-based artwork recommendation of physical paintings by studying the impact of the aforementioned features (artwork metadata, neural visual features), as well as manually-engineered visual features, such as naturalness, brightness and contrast. We implement and evaluate our method using transactional data from UGallery.com, an online artwork store. Our results show that artwork recommendations based on a hybrid combination of artist preference, curated attributes, deep neural visual features and manually-engineered visual features produce the best performance. Moreover, we discuss the trade-off between automatically obtained DNN features and manually-engineered visual features for the purpose of explainability, as well as the impact of user profile size on predictions. Our research informs the development of next-generation content-based artwork recommenders which rely on different types of data, from text to multimedia.\n
\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 2017\n \n \n (4)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n GENIUS: web server to predict local gene networks and key genes for biological functions.\n \n\n\n \n Puelma, T.; Araus, V.; Canales, J.; Vidal, E.; Cabello, J.; Soto, A.; and Gutierrez, R.\n\n\n \n\n\n\n Bioinformatics. 2017.\n \n\n\n\n
\n\n\n\n \n \n \"GENIUS:Paper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Puelma:EtAl:2017,\n  Author = {T. Puelma and V. Araus and J. Canales and E. Vidal and J. Cabello and A. Soto and R. Gutierrez},\n  Title = {GENIUS: web server to predict local gene networks and key genes for biological functions},\n  Journal = {Bioinformatics},\n  Year = {2017},\n  abstract = {GENIUS is a user-friendly web server that uses a novel machine learning algorithm to infer functional gene networks focused on specific genes and experimental conditions that are relevant to biological functions of interest. These functions may have different levels of complexity, from specific biological processes to complex traits that involve several interacting processes. GENIUS also enriches the network with new genes related to the biological function of interest, with accuracies comparable to highly discriminative Support Vector Machine methods},\nurl = {https://doi.org/10.1093/bioinformatics/btw702}\n}\n\n\n
\n
\n\n\n
\n GENIUS is a user-friendly web server that uses a novel machine learning algorithm to infer functional gene networks focused on specific genes and experimental conditions that are relevant to biological functions of interest. These functions may have different levels of complexity, from specific biological processes to complex traits that involve several interacting processes. GENIUS also enriches the network with new genes related to the biological function of interest, with accuracies comparable to highly discriminative Support Vector Machine methods\n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Sparse composition of body poses and atomic actions for human activity recognition in RGB-D videos.\n \n\n\n \n Lillo, I.; Niebles, J.; and Soto, A.\n\n\n \n\n\n\n Image and Vision Computing, 59(March): 63-75. 2017.\n \n\n\n\n
\n\n\n\n \n \n \"SparsePaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Lillo:EtAl:2017,\n  Author = {I. Lillo  and JC. Niebles and A. Soto},\n  Title = {Sparse composition of body poses and atomic actions for human activity recognition in RGB-D videos},\n  Journal = {Image and Vision Computing},\n  Volume = {59},\n  Number = {March},\n  pages={63-75},\n  Year = {2017},\n  abstract = {This paper presents an approach to recognize human activities using body poses estimated from RGB-D data. We focus on recognizing complex activities composed of sequential or simultaneous atomic actions characterized by body motions of a single actor. We tackle this problem by introducing a hierarchical compositional model that operates at three levels of abstraction. At the lowest level, geometric and motion descriptors are used to learn a dictionary of body poses. At the intermediate level, sparse compositions of these body poses are used to obtain meaningful representations for atomic human actions. Finally, at the highest level, spatial and temporal compositions of these atomic actions are used to represent complex human activities. Our results show the benefits of using a hierarchical model that exploits the sharing and composition of body poses into atomic actions, and atomic actions into activities. A quantitative evaluation using two benchmark datasets illustrates the advantages of our model to perform action and activity recognition.},\nurl={http://www.sciencedirect.com/science/article/pii/S0262885616301949}\n}\n\n\n\n
\n
\n\n\n
\n This paper presents an approach to recognize human activities using body poses estimated from RGB-D data. We focus on recognizing complex activities composed of sequential or simultaneous atomic actions characterized by body motions of a single actor. We tackle this problem by introducing a hierarchical compositional model that operates at three levels of abstraction. At the lowest level, geometric and motion descriptors are used to learn a dictionary of body poses. At the intermediate level, sparse compositions of these body poses are used to obtain meaningful representations for atomic human actions. Finally, at the highest level, spatial and temporal compositions of these atomic actions are used to represent complex human activities. Our results show the benefits of using a hierarchical model that exploits the sharing and composition of body poses into atomic actions, and atomic actions into activities. A quantitative evaluation using two benchmark datasets illustrates the advantages of our model to perform action and activity recognition.\n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Unsupervised Local Regressive Attributes for Pedestrian Re-Identification.\n \n\n\n \n Peralta, B.; Caro, L.; and Soto, A.\n\n\n \n\n\n\n In CIARP, 2017. \n \n\n\n\n
\n\n\n\n \n \n \"UnsupervisedPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Peralta:EtAl:2017,\n  Author = {B. Peralta and L. Caro and A. Soto},\n  Title = {Unsupervised Local Regressive Attributes for Pedestrian Re-Identification},\n  booktitle = {{CIARP}},\n  year = {2017},\n  abstract = {.},\nurl={http://saturno.ing.puc.cl/media/papers_alvaro/},\n}\n\n\n
\n
\n\n\n
\n .\n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Comparing Neural and Attractiveness-based Visual Features for Artwork Recommendation.\n \n\n\n \n Dominguez, V.; Messina, P.; Parra, D.; Mery, D.; Trattner, C.; and Soto, A.\n\n\n \n\n\n\n In Workshop on Deep Learning for Recommender Systems, co-located at RecSys 2017, 2017. \n \n\n\n\n
\n\n\n\n \n \n \"ComparingPaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Dominguez:EtAl:2017,\nAuthor = {V. Dominguez and P. Messina and D. Parra and D. Mery and C. Trattner and A. Soto},\nTitle = {Comparing Neural and Attractiveness-based Visual Features for Artwork Recommendation},\nYear = {2017},\nurl={https://arxiv.org/pdf/1706.07515.pdf},\nbooktitle={Workshop on Deep Learning for Recommender Systems, co-located at RecSys 2017}\n}\n\n\n\n%***********2016***************%\n\n
\n
\n\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 2016\n \n \n (3)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n A Proposal for Supervised Clustering with Dirichlet Process Using Labels.\n \n\n\n \n Peralta, B.; Caro, A.; and Soto, A.\n\n\n \n\n\n\n Pattern Recognition Letters, 80: 52-57. 2016.\n \n\n\n\n
\n\n\n\n \n \n \"APaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Peralta:EtAl:2016,\n  Author = {B. Peralta and A. Caro and A. Soto},\n  Title = {A Proposal for Supervised Clustering with Dirichlet Process Using Labels},\n  Journal = {Pattern Recognition Letters},\n  Volume = {80},\n  pages={52-57},\n  Year = {2016},\n  abstract = {Supervised clustering is an emerging area of machine learning, where the goal is to find class-uniform clusters. However, typical state-of-the-art algorithms use a fixed number of clusters. In this work, we propose a variation of a non-parametric Bayesian modeling for supervised clustering. Our approach consists of modeling the clusters as a mixture of Gaussians with the constraint of encouraging clusters of points with the same label. In order to estimate the number of clusters, we assume a-priori a countably infinite number of clusters using a variation of Dirichlet Process model over the prior distribution. In our experiments, we show that our technique typically outperforms the results of other clustering techniques.},\nurl={http://www.sciencedirect.com/science/article/pii/S0167865516300976}\n} \n\n\n
\n
\n\n\n
\n Supervised clustering is an emerging area of machine learning, where the goal is to find class-uniform clusters. However, typical state-of-the-art algorithms use a fixed number of clusters. In this work, we propose a variation of a non-parametric Bayesian modeling for supervised clustering. Our approach consists of modeling the clusters as a mixture of Gaussians with the constraint of encouraging clusters of points with the same label. In order to estimate the number of clusters, we assume a-priori a countably infinite number of clusters using a variation of Dirichlet Process model over the prior distribution. In our experiments, we show that our technique typically outperforms the results of other clustering techniques.\n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n A Hierarchical Pose-Based Approach to Complex Action Understanding Using Dictionaries of Actionlets and Motion Poselets.\n \n\n\n \n Lillo, I.; Niebles, J.; and Soto, A.\n\n\n \n\n\n\n In CVPR, 2016. \n \n\n\n\n
\n\n\n\n \n \n \"APaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Lillo:EtAl:2016,\n  Author = {I. Lillo and JC. Niebles and A. Soto},\n  Title = {A Hierarchical Pose-Based Approach to Complex Action Understanding Using Dictionaries of Actionlets and Motion Poselets},\n  booktitle = {{CVPR}},\n  year = {2016},\n  abstract = {In this paper, we introduce a new hierarchical model for human\naction\nrecognition using body joint locations.\nOur model can categorize complex actions in videos,\nand perform spatio-temporal annotations of the atomic\nactions that compose the complex action being performed.\nThat is, for each atomic action, the model generates temporal action annotations\nby\nestimating its starting and ending times, as well as,\nspatial annotations by inferring the human body parts that are involved\nin executing the action. Our model includes three key\nnovel properties:\n(i) it can be trained with no spatial\nsupervision, as it can automatically discover active body parts\nfrom temporal action annotations only;\n(ii) it jointly learns flexible representations for motion poselets and\nactionlets that encode the\nvisual variability of body parts and atomic actions;\n(iii) a mechanism to discard idle or non-informative body\nparts which increases its robustness to common pose estimation errors. We\nevaluate the performance of our method using multiple\naction recognition benchmarks. Our model consistently outperforms baselines\nand state-of-the-art action recognition methods.},\nurl={http://saturno.ing.puc.cl/media/papers_alvaro/FinalVersionActivities-CVPR-2016.pdf},\n}\n\n
\n
\n\n\n
\n In this paper, we introduce a new hierarchical model for human action recognition using body joint locations. Our model can categorize complex actions in videos, and perform spatio-temporal annotations of the atomic actions that compose the complex action being performed. That is, for each atomic action, the model generates temporal action annotations by estimating its starting and ending times, as well as, spatial annotations by inferring the human body parts that are involved in executing the action. Our model includes three key novel properties: (i) it can be trained with no spatial supervision, as it can automatically discover active body parts from temporal action annotations only; (ii) it jointly learns flexible representations for motion poselets and actionlets that encode the visual variability of body parts and atomic actions; (iii) a mechanism to discard idle or non-informative body parts which increases its robustness to common pose estimation errors. We evaluate the performance of our method using multiple action recognition benchmarks. Our model consistently outperforms baselines and state-of-the-art action recognition methods.\n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Action Recognition in Video Using Sparse Coding and Relative Features.\n \n\n\n \n Alfaro, A.; Mery, D.; and Soto, A.\n\n\n \n\n\n\n In CVPR, 2016. \n \n\n\n\n
\n\n\n\n \n \n \"ActionPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Anali:EtAl:2016,\n  author = {A. Alfaro and D. Mery and A. Soto},\n  title = {Action Recognition in Video Using Sparse Coding and Relative Features},\n  booktitle = {{CVPR}},\n  year = {2016},\n  abstract = {This work presents an approach to category-based action\nrecognition in video using sparse coding techniques. The proposed approach\nincludes two main contributions: i) A new method to handle intra-class\nvariations by decomposing\neach video into a reduced set of representative atomic action acts or\nkey-sequences, and ii) A new video descriptor, ITRA: Inter-Temporal Relational\nAct Descriptor, that exploits the power of comparative reasoning to capture\nrelative similarity relations among key-sequences. In terms of the method\nto obtain key-sequences, we introduce a loss function that, for each video,\nleads to the identification of a sparse set of representative key-frames\ncapturing both, relevant particularities arising in the input video, as well as\nrelevant generalities arising in the complete class collection. In terms of the\nmethod to obtain the ITRA descriptor, we introduce a novel scheme to quantify\nrelative intra and inter-class similarities among local temporal patterns\narising in the videos. The resulting ITRA descriptor demonstrates to be highly\neffective to discriminate among action categories. As a result, the proposed\napproach reaches remarkable action recognition performance on several popular\nbenchmark datasets, outperforming alternative state-of-the-art techniques by a\nlarge margin.},\nurl = {http://saturno.ing.puc.cl/media/papers_alvaro/FinalVersion-Anali-CVPR-2016.pdf},\n}\n\n\n%***********2015***************%\n\n
\n
\n\n\n
\n This work presents an approach to category-based action recognition in video using sparse coding techniques. The proposed approach includes two main contributions: i) A new method to handle intra-class variations by decomposing each video into a reduced set of representative atomic action acts or key-sequences, and ii) A new video descriptor, ITRA: Inter-Temporal Relational Act Descriptor, that exploits the power of comparative reasoning to capture relative similarity relations among key-sequences. In terms of the method to obtain key-sequences, we introduce a loss function that, for each video, leads to the identification of a sparse set of representative key-frames capturing both, relevant particularities arising in the input video, as well as relevant generalities arising in the complete class collection. In terms of the method to obtain the ITRA descriptor, we introduce a novel scheme to quantify relative intra and inter-class similarities among local temporal patterns arising in the videos. The resulting ITRA descriptor demonstrates to be highly effective to discriminate among action categories. As a result, the proposed approach reaches remarkable action recognition performance on several popular benchmark datasets, outperforming alternative state-of-the-art techniques by a large margin.\n
\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 2015\n \n \n (3)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n Learning Shared, Discriminative, and Compact Representations for Visual Recognition.\n \n\n\n \n Lobel, H.; Vidal, R.; and Soto, A.\n\n\n \n\n\n\n IEEE Transactions on Pattern Analysis and Machine Intelligence, 37(11). 2015.\n \n\n\n\n
\n\n\n\n \n \n \"LearningPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Lobel:EtAl:2015,\n  Author = {H. Lobel and R. Vidal and A. Soto},\n  Title = {Learning Shared, Discriminative, and Compact Representations for\nVisual Recognition},\n  Journal = {{IEEE} Transactions on Pattern Analysis and Machine Intelligence},\n  Volume = {37},\n  Number = {11},\n  Year = {2015},\n  abstract = {Dictionary-based and part-based methods are among the most popular approaches to\nvisual recognition. In both methods, a mid-level representation is built on top\nof low-level image descriptors and high-level classifiers are trained on top of\nthe mid-level representation. While earlier methods built the mid-level\nrepresentation without supervision, there is currently great interest in\nlearning both representations jointly to make the mid-level representation more\ndiscriminative. In this work we propose a new approach to visual recognition that\njointly learns a shared, discriminative, and compact mid-level representation\nand a compact high-level representation. By using a structured output learning framework, our\napproach directly handles the multiclass case at both levels of abstraction.\nMoreover, by using a group-sparse prior in the structured output learning\nframework, our approach encourages sharing of visual words and thus reduces the\nnumber of words used to represent each class. We test our proposed method on\nseveral popular benchmarks. Our results show that, by jointly learning mid- and\nhigh-level representations, and fostering the sharing of discriminative visual\nwords among target classes, we are able to achieve state-of-the-art recognition\nperformance using far less visual words than previous approaches.},\nurl = {http://saturno.ing.puc.cl/media/papers_alvaro/Hans-FINAL-PAMI-2015.pdf}\n}\n\n\n
\n
\n\n\n
\n Dictionary-based and part-based methods are among the most popular approaches to visual recognition. In both methods, a mid-level representation is built on top of low-level image descriptors and high-level classifiers are trained on top of the mid-level representation. While earlier methods built the mid-level representation without supervision, there is currently great interest in learning both representations jointly to make the mid-level representation more discriminative. In this work we propose a new approach to visual recognition that jointly learns a shared, discriminative, and compact mid-level representation and a compact high-level representation. By using a structured output learning framework, our approach directly handles the multiclass case at both levels of abstraction. Moreover, by using a group-sparse prior in the structured output learning framework, our approach encourages sharing of visual words and thus reduces the number of words used to represent each class. We test our proposed method on several popular benchmarks. Our results show that, by jointly learning mid- and high-level representations, and fostering the sharing of discriminative visual words among target classes, we are able to achieve state-of-the-art recognition performance using far less visual words than previous approaches.\n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Visual Recognition to Access and Analyze People Density and Flow Patterns in Indoor Environments.\n \n\n\n \n C. Ruz, C. P.; B. Peralta, I. L.; P. Espinace, R. G.; and B. Wendt, D. M.\n\n\n \n\n\n\n In IEEE Winter Conference on Applications of Computer Vision (WACV), 2015. \n \n\n\n\n
\n\n\n\n \n \n \"VisualPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Ruz:EtAl:2015,\n  Author = {C. Ruz, C. Pieringer, B. Peralta, I. Lillo, P. Espinace, R. Gonzalez, B. Wendt, D. Mery, A. Soto},\n  Title = {Visual Recognition to Access and Analyze People Density and Flow Patterns in Indoor Environments},\n  booktitle = {IEEE Winter Conference on Applications of Computer Vision (WACV)},\n  year = {2015},\n  abstract = {This work describes our experience developing a system to access density and flow of people in large indoor spaces using a network of RGB cameras. The\nproposed system is based on a set of overlapped and calibrated cameras. This facilitates the use of geometric constraints that\nhelp to reduce visual ambiguities. These constraints are combined with classifiers based on visual appearance to\nproduce an efficient and robust method to detect and track humans. In this work, we argue that flow\nand density of people are low level measurements that need to be complemented\nwith suitable analytic tools to bridge semantic gaps and become useful\ninformation for a target application. Consequently, we also propose a set of analytic tools that help a human user to effectively take advantage of the measurements provided by the system. Finally, we report results that demonstrate the relevance of the proposed ideas.},\nurl = {http://saturno.ing.puc.cl/media/papers_alvaro/WACV-2015-VersionFinal.pdf}\n}\n\n\n
\n
\n\n\n
\n This work describes our experience developing a system to access density and flow of people in large indoor spaces using a network of RGB cameras. The proposed system is based on a set of overlapped and calibrated cameras. This facilitates the use of geometric constraints that help to reduce visual ambiguities. These constraints are combined with classifiers based on visual appearance to produce an efficient and robust method to detect and track humans. In this work, we argue that flow and density of people are low level measurements that need to be complemented with suitable analytic tools to bridge semantic gaps and become useful information for a target application. Consequently, we also propose a set of analytic tools that help a human user to effectively take advantage of the measurements provided by the system. Finally, we report results that demonstrate the relevance of the proposed ideas.\n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Towards Improving Top-N Recommendation by Generalization of SLIM.\n \n\n\n \n Larrain, S.; Parra, D.; and Soto, A.\n\n\n \n\n\n\n In ACM RecSys, 2015. \n \n\n\n\n
\n\n\n\n \n \n \"TowardsPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{gSLIM,\n\tAuthor = {S. Larrain and D. Parra and A. Soto},\n\tBooktitle = {ACM RecSys},\n\tTitle = {Towards Improving Top-N Recommendation by Generalization of SLIM},\n\turl = {http://web.ing.puc.cl/~dparra/pdfs/Improving_SLIM_Recommendation.pdf},\n\tYear = {2015},\n        abstract = {Sparse Linear Methods (SLIM) are state-of-the-art recommendation\napproaches based on matrix factorization, which rely on a regularized\nl1-norm and l2-norm optimization –an alternative optimization\nproblem to the traditional Frobenious norm. Although\nthey have shown outstanding performance in Top-N recommendation,\nexistent works have not yet analyzed some inherent assumptions\nthat can have an important effect on the performance of these\nalgorithms. In this paper, we attempt to improve the performance\nof SLIM by proposing a generalized formulation of the aforementioned\nassumptions. Instead of directly learning a sparse representation\nof the user-item matrix, we (i) learn the latent factors’ matrix\nof the users and the items via a traditional matrix factorization approach,\nand then (ii) reconstruct the latent user or item matrix via\nprototypes which are learned using sparse coding, an alternative\nSLIM commonly used in the image processing domain. The results\nshow that by tuning the parameters of our generalized model\nwe are able to outperform SLIM in several Top-N recommendation\nexperiments conducted on two different datasets, using both nDCG\nand nDCG@10 as evaluation metrics. These preliminary results,\nalthough not conclusive, indicate a promising line of research to\nimprove the performance of SLIM recommendation.}\n}\n\n%***********2014***************%\n\n
\n
\n\n\n
\n Sparse Linear Methods (SLIM) are state-of-the-art recommendation approaches based on matrix factorization, which rely on a regularized l1-norm and l2-norm optimization –an alternative optimization problem to the traditional Frobenious norm. Although they have shown outstanding performance in Top-N recommendation, existent works have not yet analyzed some inherent assumptions that can have an important effect on the performance of these algorithms. In this paper, we attempt to improve the performance of SLIM by proposing a generalized formulation of the aforementioned assumptions. Instead of directly learning a sparse representation of the user-item matrix, we (i) learn the latent factors’ matrix of the users and the items via a traditional matrix factorization approach, and then (ii) reconstruct the latent user or item matrix via prototypes which are learned using sparse coding, an alternative SLIM commonly used in the image processing domain. The results show that by tuning the parameters of our generalized model we are able to outperform SLIM in several Top-N recommendation experiments conducted on two different datasets, using both nDCG and nDCG@10 as evaluation metrics. These preliminary results, although not conclusive, indicate a promising line of research to improve the performance of SLIM recommendation.\n
\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 2014\n \n \n (3)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n Embedded local feature selection within mixture of experts.\n \n\n\n \n Peralta, B.; and Soto, A.\n\n\n \n\n\n\n Information Sciences, 269: 176-187. 2014.\n \n\n\n\n
\n\n\n\n \n \n \"EmbeddedPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 9 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Peralta:Soto:2014,\n  Author = {B. Peralta and A. Soto},\n  Title = {Embedded local feature selection within mixture of experts},\n  Journal = {Information Sciences},\n  Volume = {269},\n  pages =  {176-187},\n  Year = {2014},\n  abstract = {A useful strategy to deal with complex classification scenarios is the divide and conquer approach. The mixture of experts (MoE) technique makes use of this strategy by jointly training a set of classifiers, or experts, that are specialized in different regions of the input space. A global model, or gate function, complements the experts by learning a function that weighs their relevance in different parts of the input space. Local feature selection appears as an attractive alternative to improve the specialization of experts and gate function, particularly, in the case of high dimensional data. In general, subsets of dimensions, or subspaces, are usually more appropriate to classify instances located in different regions of the input space. Accordingly, this work contributes with a regularized variant of MoE that incorporates an embedded process for local feature selection using L1-regularization. Experiments using artificial and real-world datasets provide evidence that the proposed method improves the classical MoE technique, in terms of accuracy and sparseness of the solution. Furthermore, our results indicate that the advantages of the proposed technique increase with the dimensionality of the data.},\nurl = {http://saturno.ing.puc.cl/media/papers_alvaro/RMoE.pdf}\n}\n\n%***********2013***************%\n\n
\n
\n\n\n
\n A useful strategy to deal with complex classification scenarios is the divide and conquer approach. The mixture of experts (MoE) technique makes use of this strategy by jointly training a set of classifiers, or experts, that are specialized in different regions of the input space. A global model, or gate function, complements the experts by learning a function that weighs their relevance in different parts of the input space. Local feature selection appears as an attractive alternative to improve the specialization of experts and gate function, particularly, in the case of high dimensional data. In general, subsets of dimensions, or subspaces, are usually more appropriate to classify instances located in different regions of the input space. Accordingly, this work contributes with a regularized variant of MoE that incorporates an embedded process for local feature selection using L1-regularization. Experiments using artificial and real-world datasets provide evidence that the proposed method improves the classical MoE technique, in terms of accuracy and sparseness of the solution. Furthermore, our results indicate that the advantages of the proposed technique increase with the dimensionality of the data.\n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Local Feature Selection Using Gaussian Process Regression.\n \n\n\n \n Pichara, K.; and Soto, A.\n\n\n \n\n\n\n Intelligent Data Analysis (IDA), 18(3). 2014.\n \n\n\n\n
\n\n\n\n \n \n \"LocalPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 22 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Pichara:EtAl:2014,\n  Author = {K. Pichara and A. Soto},\n  Title = {Local Feature Selection Using Gaussian Process Regression},\n  Journal = {Intelligent Data Analysis (IDA)},\n  Volume = {18},\n  Number = {3},\n  Year = {2014},\n  abstract = {Most feature selection algorithms determine a global subset of\nfeatures, where all\ndata instances are projected in order to improve classification accuracy. An\nattractive\nalternative solution is to adaptively find a local subset of features for each\ndata instance,\nsuch that, the classification of each instance is performed according to its own\nselective\nsubspace. This paper presents a novel application of Gaussian Processes that\nimproves\nclassification performance by learning discriminative local subsets of features\nfor each\ninstance in a dataset. Gaussian Processes are used to build for each available\nfeature\na function that estimates the discriminative power of the feature over all the\ninput\nspace. Using these functions, we are able to determine a discriminative subspace\nfor\neach possible instance by locally joining the features that present the highest\nlevels\nof discriminative power. New instances are then classified by using a K-NN\nclassifier\nthat operates in the local subspaces. Experimental results show that by using\nlocal\ndiscriminative subspaces, we are able to reach higher levels of accuracy than\nalternative\nstate-of-the-art feature selection approaches.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/Karim-IDA-2014.pdf}\n}\n\n
\n
\n\n\n
\n Most feature selection algorithms determine a global subset of features, where all data instances are projected in order to improve classification accuracy. An attractive alternative solution is to adaptively find a local subset of features for each data instance, such that, the classification of each instance is performed according to its own selective subspace. This paper presents a novel application of Gaussian Processes that improves classification performance by learning discriminative local subsets of features for each instance in a dataset. Gaussian Processes are used to build for each available feature a function that estimates the discriminative power of the feature over all the input space. Using these functions, we are able to determine a discriminative subspace for each possible instance by locally joining the features that present the highest levels of discriminative power. New instances are then classified by using a K-NN classifier that operates in the local subspaces. Experimental results show that by using local discriminative subspaces, we are able to reach higher levels of accuracy than alternative state-of-the-art feature selection approaches. \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Discriminative Hierarchical Modeling of Spatio-Temporally Composable Human Activities.\n \n\n\n \n Lillo, I.; Niebles, J.; and Soto, A.\n\n\n \n\n\n\n In CVPR, 2014. \n \n\n\n\n
\n\n\n\n \n \n \"DiscriminativePaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 33 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Lillo:EtAl:2014,\n  Author = {I. Lillo and JC. Niebles and A. Soto},\n  Title = {Discriminative Hierarchical Modeling of Spatio-Temporally Composable Human Activities},\n  booktitle = {{CVPR}},\n  year = {2014},\n  abstract = {This paper proposes a framework for recognizing complex human activities\nin videos. Our method describes human activities in a hierarchical discriminative model\nthat operates at three semantic levels. At the lower level, body poses are encoded in a representative but discriminative\npose dictionary. At the intermediate level,\nencoded poses span a space where simple human actions\nare composed. At the highest level, our model captures temporal and\nspatial compositions of actions into complex human activities.\nOur human activity classifier simultaneously models which body parts\nare relevant to the action of interest as well as their\nappearance and composition using a discriminative approach.\nBy formulating model learning in a max-margin framework,\nour approach achieves powerful multi-class discrimination while\nproviding useful annotations at the intermediate semantic level.\nWe show how our hierarchical compositional model provides natural\nhandling of occlusions.\nTo evaluate the effectiveness of our proposed framework, we introduce\na new dataset of composed human activities. We provide empirical evidence\nthat our method achieves state-of-the-art activity classification performance\non several benchmark datasets.},\nurl = {http://saturno.ing.puc.cl/media/papers_alvaro/activities-CVPR-14.pdf}\n}\n\n
\n
\n\n\n
\n This paper proposes a framework for recognizing complex human activities in videos. Our method describes human activities in a hierarchical discriminative model that operates at three semantic levels. At the lower level, body poses are encoded in a representative but discriminative pose dictionary. At the intermediate level, encoded poses span a space where simple human actions are composed. At the highest level, our model captures temporal and spatial compositions of actions into complex human activities. Our human activity classifier simultaneously models which body parts are relevant to the action of interest as well as their appearance and composition using a discriminative approach. By formulating model learning in a max-margin framework, our approach achieves powerful multi-class discrimination while providing useful annotations at the intermediate semantic level. We show how our hierarchical compositional model provides natural handling of occlusions. To evaluate the effectiveness of our proposed framework, we introduce a new dataset of composed human activities. We provide empirical evidence that our method achieves state-of-the-art activity classification performance on several benchmark datasets.\n
\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 2013\n \n \n (6)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n Hierarchical Joint Max-Margin Learning of Mid and Top Level Representations for Visual Recognition.\n \n\n\n \n Lobel, H.; Vidal, R.; and Soto, A.\n\n\n \n\n\n\n In ICCV, 2013. \n \n\n\n\n
\n\n\n\n \n \n \"HierarchicalPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 6 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Lobel-a:EtAl:2013,\n  Author = {H. Lobel and R. Vidal and A. Soto},\n  Title = {Hierarchical Joint Max-Margin Learning of Mid and Top Level Representations for Visual Recognition},\n  booktitle = {{ICCV}},\n  year = {2013},\n  abstract = {Currently, Bag-of-Visual-Words (BoVW) and\npart-based methods are the most popular approaches for visual recognition. In\nboth cases, a mid-level representation is build on top of low level image\ndescriptors while top levels classifiers use this mid-level\nrepresentation to achieve visual recognition. While in current part-based\napproaches, mid and top level representations are usually jointly trained, this\nis not the usual case for BoVW schemes. A main reason is the complex data\nassociation problem associated to the larger size of the visual dictionary\nusually needed by BoVW approaches at the mid-level layer. As a further\nobservation, typical solutions based on BoVW and part-based representations are\nusually limited to binary classification problems, a strategy that ignores\nrelevant correlations among classes. In this work we propose a novel\nhierarchical approach for visual recognition that, in the context of a BoVW\nscheme, jointly learns suitable mid and top level representations. Furthermore,\nusing a max-margin learning framework, the proposed approach directly handles\nthe multiclass case at both levels of abstraction. We test our proposed method\nusing several popular benchmarks datasets. As our main result, we demonstrate\nthat by coupling learning\nof mid and top level representations, the proposed approach fosters\nsharing of discriminativity words among target\nclasses, being able to achieve\nstate-of-the-art recognition performance using far less visual words than\nprevious approaches.},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/finalHans-ICCV-13.pdf}\n}\n\n\n
\n
\n\n\n
\n Currently, Bag-of-Visual-Words (BoVW) and part-based methods are the most popular approaches for visual recognition. In both cases, a mid-level representation is build on top of low level image descriptors while top levels classifiers use this mid-level representation to achieve visual recognition. While in current part-based approaches, mid and top level representations are usually jointly trained, this is not the usual case for BoVW schemes. A main reason is the complex data association problem associated to the larger size of the visual dictionary usually needed by BoVW approaches at the mid-level layer. As a further observation, typical solutions based on BoVW and part-based representations are usually limited to binary classification problems, a strategy that ignores relevant correlations among classes. In this work we propose a novel hierarchical approach for visual recognition that, in the context of a BoVW scheme, jointly learns suitable mid and top level representations. Furthermore, using a max-margin learning framework, the proposed approach directly handles the multiclass case at both levels of abstraction. We test our proposed method using several popular benchmarks datasets. As our main result, we demonstrate that by coupling learning of mid and top level representations, the proposed approach fosters sharing of discriminativity words among target classes, being able to achieve state-of-the-art recognition performance using far less visual words than previous approaches.\n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Human Action Recognition from Inter-Temporal Dictionaries of Key-Sequences.\n \n\n\n \n Alfaro, A.; Mery, D.; and Soto, A.\n\n\n \n\n\n\n In 6th Pacific-Rim Symposium on Image and Video Technology, PSIVT, 2013. \n \n\n\n\n
\n\n\n\n \n \n \"HumanPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 14 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Alfaro:EtAl:2013,\n  Author = { A. Alfaro and D. Mery and A. Soto},\n  Title = {Human Action Recognition from Inter-Temporal Dictionaries of\nKey-Sequences},\n  booktitle = {6th Pacific-Rim Symposium on Image and Video Technology, PSIVT},\n  year = {2013},\n  abstract = {This paper addresses the human action recognition in video\nby proposing a method based on three main processing steps. First, we\ntackle problems related to intraclass variations and differences in video\nlengths. We achieve this by reducing an input video to a set of key-sequences\nthat represent atomic meaningful acts of each action class.\nSecond, we use sparse coding techniques to learn a representation for\neach key-sequence. We then join these representations still preserving\ninformation about temporal relationships. We believe that this is a key\nstep of our approach because it provides not only a suitable shared rep\nresentation to characterize atomic acts, but it also encodes global tem\nporal consistency among these acts. Accordingly, we call this represen\ntation inter-temporal acts descriptor. Third, we use this representation\nand sparse coding techniques to classify new videos. Finally, we show\nthat, our approach outperforms several state-of-the-art methods when is\ntested using common benchmarks.},\nurl = {http://saturno.ing.puc.cl/media/papers_alvaro/Anali-PSIVT-13.pdf}\n}\n\n\n
\n
\n\n\n
\n This paper addresses the human action recognition in video by proposing a method based on three main processing steps. First, we tackle problems related to intraclass variations and differences in video lengths. We achieve this by reducing an input video to a set of key-sequences that represent atomic meaningful acts of each action class. Second, we use sparse coding techniques to learn a representation for each key-sequence. We then join these representations still preserving information about temporal relationships. We believe that this is a key step of our approach because it provides not only a suitable shared rep resentation to characterize atomic acts, but it also encodes global tem poral consistency among these acts. Accordingly, we call this represen tation inter-temporal acts descriptor. Third, we use this representation and sparse coding techniques to classify new videos. Finally, we show that, our approach outperforms several state-of-the-art methods when is tested using common benchmarks.\n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Joint Dictionary and Classifier learning for Categorization of Images using a Max-margin Framework.\n \n\n\n \n Lobel, H.; Vidal, R.; Mery, D.; and Soto., A.\n\n\n \n\n\n\n In 6th Pacific-Rim Symposium on Image and Video Technology, PSIVT, 2013. \n \n\n\n\n
\n\n\n\n \n \n \"JointPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 7 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Lobel-b:EtAl:2013,\n  Author = {H. Lobel and R. Vidal and D. Mery and A. Soto.},\n  Title = {Joint Dictionary and Classifier learning for Categorization of Images\nusing a Max-margin Framework},\n  booktitle = {6th Pacific-Rim Symposium on Image and Video Technology, PSIVT},\n  year = {2013},\n    abstract = {The Bag-of-Visual-Words (BoVW) model is a popular approach for\nvisual recognition. Used successfully in many different tasks, simplicity and good\nperformance are the main reasons for its popularity. The central aspect of this\nmodel, the visual dictionary, is used to build mid-level representations based on\nlow level image descriptors. Classifiers are then trained using these mid-level\nrepresentations to perform categorization. While most works based on BoVW\nmodels have been focused on learning a suitable dictionary or on proposing a\nsuitable pooling strategy, little effort has been devoted to explore and improve\nthe coupling between the dictionary and the top-level classifiers, in order to gen-\nerate more discriminative models. This problem can be highly complex due to\nthe large dictionary size usually needed by these methods. Also, most BoVW\nbased systems usually perform multiclass categorization using a one-vs-all strat-\negy, ignoring relevant correlations among classes. To tackle the previous issues,\nwe propose a novel approach that jointly learns dictionary words and a proper top-\nlevel multiclass classifier. We use a max-margin learning framework to minimize\na regularized energy formulation, allowing us to propagate labeled information\nto guide the commonly unsupervised dictionary learning process. As a result we\nproduce a dictionary that is more compact and discriminative. We test our method\non several popular datasets, where we demonstrate that our joint optimization\nstrategy induces a word sharing behavior among the target classes, being able\nto achieve state-of-the-art performance using far less visual words than previous\napproaches.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/Hans-PSIVT-13.pdf}\n}\n\n\n
\n
\n\n\n
\n The Bag-of-Visual-Words (BoVW) model is a popular approach for visual recognition. Used successfully in many different tasks, simplicity and good performance are the main reasons for its popularity. The central aspect of this model, the visual dictionary, is used to build mid-level representations based on low level image descriptors. Classifiers are then trained using these mid-level representations to perform categorization. While most works based on BoVW models have been focused on learning a suitable dictionary or on proposing a suitable pooling strategy, little effort has been devoted to explore and improve the coupling between the dictionary and the top-level classifiers, in order to gen- erate more discriminative models. This problem can be highly complex due to the large dictionary size usually needed by these methods. Also, most BoVW based systems usually perform multiclass categorization using a one-vs-all strat- egy, ignoring relevant correlations among classes. To tackle the previous issues, we propose a novel approach that jointly learns dictionary words and a proper top- level multiclass classifier. We use a max-margin learning framework to minimize a regularized energy formulation, allowing us to propagate labeled information to guide the commonly unsupervised dictionary learning process. As a result we produce a dictionary that is more compact and discriminative. We test our method on several popular datasets, where we demonstrate that our joint optimization strategy induces a word sharing behavior among the target classes, being able to achieve state-of-the-art performance using far less visual words than previous approaches. \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Enhancing K-Means Using Class Labels.\n \n\n\n \n Peralta, B.; Espinace, P.; and Soto, A.\n\n\n \n\n\n\n Intelligent Data Analysis (IDA), 17(6): 1023-1039. 2013.\n \n\n\n\n
\n\n\n\n \n \n \"EnhancingPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 9 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Peralta:EtAl:2013,\n  Author = {B. Peralta and P. Espinace and A. Soto},\n  Title = {Enhancing K-Means Using Class Labels},\n  Journal = {Intelligent Data Analysis (IDA)},\n  Volume = {17},\n  Number = {6},\n  pages = {1023-1039},\n  Year = {2013},\n  abstract = {Clustering is a relevant problem in machine learning where the\nmain goal is\nto locate meaningful partitions of unlabeled data. In the case of labeled data,\na\nrelated problem is supervised clustering, where the objective is to locate\nclass-\nuniform clusters. Most current approaches to supervised clustering optimize a\nscore related to cluster purity with respect to class labels. In particular, we\npresent\nLabeled K-Means (LK-Means), an algorithm for supervised clustering based on a\nvariant of K-Means that incorporates information about class labels. LK-Means\nreplaces the classical cost function of K-Means by a convex combination of the\njoint cost associated to: (i) A discriminative score based on class labels, and\n(ii)\nA generative score based on a traditional metric for unsupervised clustering. We\ntest the performance of LK-Means using standard real datasets and an\napplication for object recognition. Moreover, we also compare its performance\nagainst\nclassical K-Means and a popular K-Medoids-based supervised clustering method.\nOur experiments show that, in most cases, LK-Means outperforms the alternative\ntechniques by a considerable margin. Furthermore, LK-Means presents execution\ntimes considerably lower than the alternative supervised clustering method under\nevaluation.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/supClustering.pdf}\n}\n\n
\n
\n\n\n
\n Clustering is a relevant problem in machine learning where the main goal is to locate meaningful partitions of unlabeled data. In the case of labeled data, a related problem is supervised clustering, where the objective is to locate class- uniform clusters. Most current approaches to supervised clustering optimize a score related to cluster purity with respect to class labels. In particular, we present Labeled K-Means (LK-Means), an algorithm for supervised clustering based on a variant of K-Means that incorporates information about class labels. LK-Means replaces the classical cost function of K-Means by a convex combination of the joint cost associated to: (i) A discriminative score based on class labels, and (ii) A generative score based on a traditional metric for unsupervised clustering. We test the performance of LK-Means using standard real datasets and an application for object recognition. Moreover, we also compare its performance against classical K-Means and a popular K-Medoids-based supervised clustering method. Our experiments show that, in most cases, LK-Means outperforms the alternative techniques by a considerable margin. Furthermore, LK-Means presents execution times considerably lower than the alternative supervised clustering method under evaluation. \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Indoor Scene Recognition by a Mobile Robot Through Adaptive Object Detection.\n \n\n\n \n Espinace, P.; Kollar, T.; Roy, N.; and Soto, A.\n\n\n \n\n\n\n Robotics and Autonomous Systems, 61(9). 2013.\n \n\n\n\n
\n\n\n\n \n \n \"IndoorPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 5 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Espinace:EtAl:2013,\n  Author = {P. Espinace and T. Kollar and N. Roy and A. Soto},\n  Title = {Indoor Scene Recognition by a Mobile Robot Through Adaptive Object\nDetection},\n  Journal = {Robotics and Autonomous Systems},\n  Volume = {61},\n  Number = {9},\n  Year = {2013},\n  abstract = {Mobile Robotics has achieved notably progress, however, to\nincrease the\ncomplexity of the tasks that mobile robots can perform in natural environments,\nwe need to provide them with a greater semantic understanding of\ntheir surrounding. In particular, identifying indoor scenes, such as an office\nor a kitchen, is a highly valuable perceptual ability for an indoor mobile\nrobot, and in this paper we propose a new technique to achieve this goal. As\na distinguishing feature, we use common objects, such as doors or furnitures,\nas a key intermediate representation to recognize indoor scenes. We frame\nour method as a generative probabilistic hierarchical model, where we use\nobject category classifiers to associate low-level visual features to objects,\nand contextual relations to associate objects to scenes. The inherent seman-\ntic interpretation of common objects allows us to use rich sources of online\ndata to populate the probabilistic terms of our model. In contrast to alterna-\ntive computer vision based methods, we boost performance by exploiting the\nembedded and dynamic nature of a mobile robot. In particular, we increase\ndetection accuracy and efficiency by using a 3D range sensor that allows us\nto implement a focus of attention mechanism based on geometric and struc-\ntural information. Furthermore, we use concepts from information theory\nto propose an adaptive scheme that limits computational load by selectively\nguiding the search for informative objects. The operation of this scheme is\nfacilitated by the dynamic nature of a mobile robot that is constantly changing\nits field of view. We test our approach using real data captured by a mo-\nbile robot navigating in office and home environments. Our results indicate\nthat the proposed approach outperforms several state-of-the-art techniques\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/Final-RAS-2013.pdf}\n}\n\n\n
\n
\n\n\n
\n Mobile Robotics has achieved notably progress, however, to increase the complexity of the tasks that mobile robots can perform in natural environments, we need to provide them with a greater semantic understanding of their surrounding. In particular, identifying indoor scenes, such as an office or a kitchen, is a highly valuable perceptual ability for an indoor mobile robot, and in this paper we propose a new technique to achieve this goal. As a distinguishing feature, we use common objects, such as doors or furnitures, as a key intermediate representation to recognize indoor scenes. We frame our method as a generative probabilistic hierarchical model, where we use object category classifiers to associate low-level visual features to objects, and contextual relations to associate objects to scenes. The inherent seman- tic interpretation of common objects allows us to use rich sources of online data to populate the probabilistic terms of our model. In contrast to alterna- tive computer vision based methods, we boost performance by exploiting the embedded and dynamic nature of a mobile robot. In particular, we increase detection accuracy and efficiency by using a 3D range sensor that allows us to implement a focus of attention mechanism based on geometric and struc- tural information. Furthermore, we use concepts from information theory to propose an adaptive scheme that limits computational load by selectively guiding the search for informative objects. The operation of this scheme is facilitated by the dynamic nature of a mobile robot that is constantly changing its field of view. We test our approach using real data captured by a mo- bile robot navigating in office and home environments. Our results indicate that the proposed approach outperforms several state-of-the-art techniques \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Automated Design of a Computer Vision System for Food Quality Evaluation.\n \n\n\n \n Mery, D.; Pedreschi, F.; and Soto, A.\n\n\n \n\n\n\n Food and Bioprocess Technology, 6(8): 2093-2108. 2013.\n \n\n\n\n
\n\n\n\n \n \n \"AutomatedPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 3 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Mery:EtAl:2013,\n  Author = {D. Mery and F. Pedreschi and A. Soto},\n  Title = {Automated Design of a Computer Vision System for Food Quality\nEvaluation},\n  Journal = {Food and Bioprocess Technology},\n  Volume = {6},\n  number = {8},\n  pages = {2093-2108},\n  Year = {2013},\n  abstract = {Considerable research efforts in computer classifiers for a given\napplication avoiding the classical\nvision applied to food quality evaluation have been trial and error framework\ncommonly used by human\ndeveloped in the last years; however, they have been designers. The key idea of\nthe proposed framework\nconcentrated on using or developing tailored methods is to\nselect—automatically—from a large set of fea-\nbased on visual features that are able to solve a specific tures and a bank of\nclassifiers those features and clas-\ntask. Nevertheless, today’s computer capabilities are sifiers that achieve the\nhighest performance. We tested\n  giving us new ways to solve complex computer vision our framework on eight\ndifferent food quality evalua-\n  problems. In particular, a new paradigm on machine tion problems yielding a\nclassification performance of\n  learning techniques has emerged posing the task of 95 % or more in every case.\nThe proposed framework\n  recognizing visual patterns as a search problem based was implemented as a\nMatlab Toolbox available for\n  on training data and a hypothesis space composed by noncommercial purposes.\n  visual features and suitable classifiers. Furthermore,\n  now we are able to extract, process, and test in the same\n  time more image features and classifiers than before.\n  Thus, we propose a general framework that designs a\n  computer vision system automatically, i.e., it finds—\n    without human interaction—the features and the classifiers for a given\napplication avoiding the classical\ntrial and error framework commonly used by human\ndesigners. The key idea of the proposed framework\nis to select—automatically—from a large set of fea-\n   tures and a bank of classifiers those features and clas-\n  sifiers that achieve the highest performance. We tested\n our framework on eight different food quality evalua-\ntion problems yielding a classification performance of\n95% or more in every case. The proposed framework\nwas implemented as a Matlab Toolbox available for\nnoncommercial purposes.\n\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/Food-Mery-2012.pdf}\n}\n\n\n%***********2012***************%\n\n\n
\n
\n\n\n
\n Considerable research efforts in computer classifiers for a given application avoiding the classical vision applied to food quality evaluation have been trial and error framework commonly used by human developed in the last years; however, they have been designers. The key idea of the proposed framework concentrated on using or developing tailored methods is to select—automatically—from a large set of fea- based on visual features that are able to solve a specific tures and a bank of classifiers those features and clas- task. Nevertheless, today’s computer capabilities are sifiers that achieve the highest performance. We tested giving us new ways to solve complex computer vision our framework on eight different food quality evalua- problems. In particular, a new paradigm on machine tion problems yielding a classification performance of learning techniques has emerged posing the task of 95 % or more in every case. The proposed framework recognizing visual patterns as a search problem based was implemented as a Matlab Toolbox available for on training data and a hypothesis space composed by noncommercial purposes. visual features and suitable classifiers. Furthermore, now we are able to extract, process, and test in the same time more image features and classifiers than before. Thus, we propose a general framework that designs a computer vision system automatically, i.e., it finds— without human interaction—the features and the classifiers for a given application avoiding the classical trial and error framework commonly used by human designers. The key idea of the proposed framework is to select—automatically—from a large set of fea- tures and a bank of classifiers those features and clas- sifiers that achieve the highest performance. We tested our framework on eight different food quality evalua- tion problems yielding a classification performance of 95% or more in every case. The proposed framework was implemented as a Matlab Toolbox available for noncommercial purposes. \n
\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 2012\n \n \n (3)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n Discriminative local subspaces in gene expression data for effective gene function prediction.\n \n\n\n \n Puelma, T.; Gutierrez, R.; and Soto, A.\n\n\n \n\n\n\n Bioinformatics, 28(17): 2256-64. 2012.\n \n\n\n\n
\n\n\n\n \n \n \"DiscriminativePaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 3 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Puelma:EtAl:2012,\n  Author = {T. Puelma and R. Gutierrez and A. Soto},\n  Title = {Discriminative local subspaces in gene expression data for effective\ngene function prediction},\n  Journal = {Bioinformatics},\n  Volume = {28},\n  Number = {17},\n  pages={2256-64},\n  Year = {2012},\n  abstract = {Motivation: Massive amounts of genome-wide gene expression data\nhave become available, motivating the development of computatio-\nnal approaches that leverage this information to predict gene func-\ntion. Among successful approaches, supervised machine learning\nmethods, such as Support Vector Machines, have shown superior\nprediction accuracy. However, these methods lack the simple biologi-\ncal intuition provided by coexpression networks, limiting their practical\nusefulness.\nResults: In this work we present Discriminative Local Subspaces\n(DLS), a novel method that combines supervised machine learning\nand coexpression techniques with the goal of systematically predict\ngenes involved in specific biological processes of interest. Unlike tra-\nditional coexpression networks, DLS uses the knowledge available in\nGene Ontology (GO) to generate informative training sets that guide\nthe discovery of expression signatures: expression patterns that are\ndiscriminative for genes involved in the biological process of interest.\nBy linking genes coexpressed with these signatures, DLS is able\nto construct a discriminative coexpression network that links both,\nknown and previously uncharacterized genes, for the selected bio-\nlogical process. This paper focuses on the algorithm behind DLS and\nshows its predictive power using an Arabidopsis thaliana dataset and\na representative set of 101 GO-terms from the Biological Process\nOntology. Our results show that DLS has a superior average accuracy\nthan both, Support Vector Machines and Coexpression Networks.\nThus, DLS is able to provide the prediction accuracy of supervised\nlearning methods, while maintaining the intuitive understanding of\ncoexpression networks.\nAvailability and Implementation: A MATLAB R implementation\nof DLS is available at http://virtualplant.bio.puc.cl/\ncgi-bin/Lab/tools.cgi.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/DLS_Revised_Paper.pdf}\n}\n\n\n
\n
\n\n\n
\n Motivation: Massive amounts of genome-wide gene expression data have become available, motivating the development of computatio- nal approaches that leverage this information to predict gene func- tion. Among successful approaches, supervised machine learning methods, such as Support Vector Machines, have shown superior prediction accuracy. However, these methods lack the simple biologi- cal intuition provided by coexpression networks, limiting their practical usefulness. Results: In this work we present Discriminative Local Subspaces (DLS), a novel method that combines supervised machine learning and coexpression techniques with the goal of systematically predict genes involved in specific biological processes of interest. Unlike tra- ditional coexpression networks, DLS uses the knowledge available in Gene Ontology (GO) to generate informative training sets that guide the discovery of expression signatures: expression patterns that are discriminative for genes involved in the biological process of interest. By linking genes coexpressed with these signatures, DLS is able to construct a discriminative coexpression network that links both, known and previously uncharacterized genes, for the selected bio- logical process. This paper focuses on the algorithm behind DLS and shows its predictive power using an Arabidopsis thaliana dataset and a representative set of 101 GO-terms from the Biological Process Ontology. Our results show that DLS has a superior average accuracy than both, Support Vector Machines and Coexpression Networks. Thus, DLS is able to provide the prediction accuracy of supervised learning methods, while maintaining the intuitive understanding of coexpression networks. Availability and Implementation: A MATLAB R implementation of DLS is available at http://virtualplant.bio.puc.cl/ cgi-bin/Lab/tools.cgi. \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Adaptive hierarchical contexts for object recognition with conditional mixture of trees.\n \n\n\n \n Peralta, B.; Espinace, P.; and Soto, A.\n\n\n \n\n\n\n In BMVC, 2012. \n \n\n\n\n
\n\n\n\n \n \n \"AdaptivePaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 3 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Peralta:EtAl:2012,\n  Author = {B. Peralta and P. Espinace and A. Soto},\n  Title = {Adaptive hierarchical contexts for object recognition with\nconditional mixture of trees},\n  booktitle = {{BMVC}},\n  year = {2012},\n    abstract = {Robust category-level object recognition is currently a major goal for the computer\nvision community. Intra-class and pose variations, as well as, background clutter and\npartial occlusions are some of the main difficulties to achieve this goal. Contextual in-\nformation, in the form of object co-occurrences and spatial constraints, has been suc-\ncessfully applied to improve object recognition performance, however, previous work\nconsiders only fixed contextual relations that do not depend of the type of scene under\ninspection. In this work, we present a method that learns adaptive conditional relation-\nships that depend on the type of scene being analyzed. In particular, we propose a model\nbased on a conditional mixture of trees that is able to capture contextual relationships\namong objects using global information about a scene. Our experiments show that the\nadaptive specialization of contextual relationships improves object recognition accuracy\noutperforming previous state-of-the-art approaches.\n},\n  url = {FinalBMVC-12.pdf}\n}\n\n\n
\n
\n\n\n
\n Robust category-level object recognition is currently a major goal for the computer vision community. Intra-class and pose variations, as well as, background clutter and partial occlusions are some of the main difficulties to achieve this goal. Contextual in- formation, in the form of object co-occurrences and spatial constraints, has been suc- cessfully applied to improve object recognition performance, however, previous work considers only fixed contextual relations that do not depend of the type of scene under inspection. In this work, we present a method that learns adaptive conditional relation- ships that depend on the type of scene being analyzed. In particular, we propose a model based on a conditional mixture of trees that is able to capture contextual relationships among objects using global information about a scene. Our experiments show that the adaptive specialization of contextual relationships improves object recognition accuracy outperforming previous state-of-the-art approaches. \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Indoor Mobile Robotics at Grima, PUC.\n \n\n\n \n L. Caro, J. C.; P. Espinace, D. M.; R. Mitnik, S. M.; S. Pszszfolkowski, D. L.; A. Araneda, D. M.; and M. Torres, A. S.\n\n\n \n\n\n\n Journal of Intelligent and Robotic Systems, 66(1-2): 151-165. 2012.\n \n\n\n\n
\n\n\n\n \n \n \"IndoorPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Caro:EtAl:2012,\n  Author = {L. Caro, J. Correa, P. Espinace, D. Maturana, R. Mitnik, S.\nMontabone, S. Pszszfolkowski, D. Langdon, A. Araneda, D. Mery, M. Torres, A.\nSoto},\n  Title = {Indoor Mobile Robotics at Grima, PUC},\n  Journal = {Journal of Intelligent and Robotic Systems},\n  Volume = {66},\n  pages={151-165},\n  Number = {1-2},\n  Year = {2012},\n    abstract = {This paper describes the main activities\nand achievements of our research group on Ma-\nchine Intelligence and Robotics (Grima) at the\nComputer Science Department, Pontificia Uni-\nversidad Catolica de Chile (PUC). Since 2002,\nwe have been developing an active research in\nthe area of indoor autonomous social robots.\nOur main focus has been the cognitive side of\nRobotics, where we have developed algorithms\nfor autonomous navigation using wheeled robots,\nscene recognition using vision and 3D range sen-\nsors, and social behaviors using Markov Deci-\nsion Processes, among others. As a distinguish-\ning feature, in our research we have followed\na probabilistic approach, deeply rooted in ma-\nchine learning and Bayesian statistical techniques.\nAmong our main achievements are an increasing\nlist of publications in main Robotics conference\nand journals, and the consolidation of a research\ngroup with more than 25 people among full-\ntime professors, visiting researchers, and graduate\nstudents.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/Latam-2012.pdf}\n}\n\n%***********2011***************%\n\n
\n
\n\n\n
\n This paper describes the main activities and achievements of our research group on Ma- chine Intelligence and Robotics (Grima) at the Computer Science Department, Pontificia Uni- versidad Catolica de Chile (PUC). Since 2002, we have been developing an active research in the area of indoor autonomous social robots. Our main focus has been the cognitive side of Robotics, where we have developed algorithms for autonomous navigation using wheeled robots, scene recognition using vision and 3D range sen- sors, and social behaviors using Markov Deci- sion Processes, among others. As a distinguish- ing feature, in our research we have followed a probabilistic approach, deeply rooted in ma- chine learning and Bayesian statistical techniques. Among our main achievements are an increasing list of publications in main Robotics conference and journals, and the consolidation of a research group with more than 25 people among full- time professors, visiting researchers, and graduate students. \n
\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 2011\n \n \n (4)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n Learning Discriminative Local Binary Patterns for Face Recognition.\n \n\n\n \n Maturana, D.; Mery, D.; and Soto, A.\n\n\n \n\n\n\n In 9th IEEE International Conference on Automatic Face and Gesture Recognition (FG), 2011. \n \n\n\n\n
\n\n\n\n \n \n \"LearningPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 16 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Maturana:EtAl:2011,\n  Author = {D. Maturana and D. Mery and A. Soto},\n  Title = {Learning Discriminative Local Binary Patterns for Face Recognition},\n  booktitle = {9th IEEE International Conference on Automatic Face and Gesture\nRecognition (FG)},\n  year = {2011},\n    abstract = {Histograms of Local Binary Patterns (LBPs) and\nvariations thereof are a popular local visual descriptor for\nface recognition. So far, most variations of LBP are designed\nby hand or are learned with non-supervised methods. In this\nwork we propose a simple method to learn discriminative LBPs\nin a supervised manner. The method represents an LBP-like\ndescriptor as a set of pixel comparisons within a neighborhood\nand heuristically seeks for a set of pixel comparisons so as to\nmaximize a Fisher separability criterion for the resulting his-\ntograms. Tests on standard face recognition datasets show that\nthis method can create compact yet discriminative descriptors.},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/FG-2011.pdf}\n}\n\n\n
\n
\n\n\n
\n Histograms of Local Binary Patterns (LBPs) and variations thereof are a popular local visual descriptor for face recognition. So far, most variations of LBP are designed by hand or are learned with non-supervised methods. In this work we propose a simple method to learn discriminative LBPs in a supervised manner. The method represents an LBP-like descriptor as a set of pixel comparisons within a neighborhood and heuristically seeks for a set of pixel comparisons so as to maximize a Fisher separability criterion for the resulting his- tograms. Tests on standard face recognition datasets show that this method can create compact yet discriminative descriptors.\n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Automated Fish Bone Detection using X-ray Testing.\n \n\n\n \n Mery, D.; Lillo, I.; Loebel, H.; V. Riffo, A. S.; Cipriano, A.; and Aguilera, J.\n\n\n \n\n\n\n Journal of Food Engineering, 105(3): 485-492. 2011.\n \n\n\n\n
\n\n\n\n \n \n \"AutomatedPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Mery:EtAl:2011,\n  Author = {D. Mery and I. Lillo and H. Loebel and V. Riffo, A. Soto and  A.\nCipriano and JM. Aguilera},\n  Title = {Automated Fish Bone Detection using X-ray Testing},\n  journal = {Journal of Food Engineering},\n  volume = {105},\n  number = {3},\n  pages={485-492},\n  year = {2011},\n    abstract = {In countries where fish is often consumed, fish bones are some of the most frequently ingested foreign bodies encountered in foods.\nIn the production of fish fillets, fish bone detection is performed by human inspection using their sense of touch and vision which\ncan lead to misclassification. Effective detection of fish bones in the quality control process would help avoid this problem. For this\nreason, an X-ray machine vision approach to automatically detect fish bones in fish fillets was developed. This paper describes our\napproach and the corresponding experiments with salmon and trout fillets. In the experiments, salmon X-ray images using 10×10\npixels detection windows and 24 intensity features (selected from 279 features) were analyzed. The methodology was validated\nusing representative fish bones and trouts provided by a salmon industry and yielded a detection performance of 99%. We believe\nthat the proposed approach opens new possibilities in the field of automated visual inspection of salmon, trout and other similar\nfish.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/2011-JFoodEng-SalmonX.pdf},\n}\n\n\n
\n
\n\n\n
\n In countries where fish is often consumed, fish bones are some of the most frequently ingested foreign bodies encountered in foods. In the production of fish fillets, fish bone detection is performed by human inspection using their sense of touch and vision which can lead to misclassification. Effective detection of fish bones in the quality control process would help avoid this problem. For this reason, an X-ray machine vision approach to automatically detect fish bones in fish fillets was developed. This paper describes our approach and the corresponding experiments with salmon and trout fillets. In the experiments, salmon X-ray images using 10×10 pixels detection windows and 24 intensity features (selected from 279 features) were analyzed. The methodology was validated using representative fish bones and trouts provided by a salmon industry and yielded a detection performance of 99%. We believe that the proposed approach opens new possibilities in the field of automated visual inspection of salmon, trout and other similar fish. \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Active Learning and Subspace Clustering for Anomaly Detection.\n \n\n\n \n Pichara, K.; and Soto, A.\n\n\n \n\n\n\n Intelligent Data Analysis (IDA), 15(2): 151-171. 2011.\n \n\n\n\n
\n\n\n\n \n \n \"ActivePaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 1 download\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Pichara:EtAl:2011,\n  Author = {K. Pichara and A. Soto},\n  Title = {Active Learning and Subspace Clustering for Anomaly Detection},\n  journal = { Intelligent Data Analysis (IDA)},\n  volume = {15},\n  number = {2},\n  pages={151-171},\n  year = {2011},\n    abstract = {Today, anomaly detection is a highly valuable application in the analysis of current huge datasets. Insurance\ncompanies, banks and many manufacturing industries need systems to help humans to detect anomalies in their daily information.\nIn general, anomalies are a very small fraction of the data, therefore their detection is not an easy task. Usually real sources of\nan anomaly are given by specific values expressed on selective dimensions of datasets, furthermore, many anomalies are not\nreally interesting for humans, due to the fact that interestingness of anomalies is categorized subjectively by the human user.\nIn this paper we propose a new semi-supervised algorithm that actively learns to detect relevant anomalies by interacting with\nan expert user in order to obtain semantic information about user preferences. Our approach is based on 3 main steps. First, a\nBayes network identifies an initial set of candidate anomalies. Afterwards, a subspace clustering technique identifies relevant\nsubsets of dimensions. Finally, a probabilistic active learning scheme, based on properties of Dirichlet distribution, uses the\nfeedback from an expert user to efficiently search for relevant anomalies. Our results, using synthetic and real datasets, indicate\nthat, under noisy data and anomalies presenting regular patterns, our approach correctly identifies relevant anomalies.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/IDA-2011.pdf}\n}\n\n\n
\n
\n\n\n
\n Today, anomaly detection is a highly valuable application in the analysis of current huge datasets. Insurance companies, banks and many manufacturing industries need systems to help humans to detect anomalies in their daily information. In general, anomalies are a very small fraction of the data, therefore their detection is not an easy task. Usually real sources of an anomaly are given by specific values expressed on selective dimensions of datasets, furthermore, many anomalies are not really interesting for humans, due to the fact that interestingness of anomalies is categorized subjectively by the human user. In this paper we propose a new semi-supervised algorithm that actively learns to detect relevant anomalies by interacting with an expert user in order to obtain semantic information about user preferences. Our approach is based on 3 main steps. First, a Bayes network identifies an initial set of candidate anomalies. Afterwards, a subspace clustering technique identifies relevant subsets of dimensions. Finally, a probabilistic active learning scheme, based on properties of Dirichlet distribution, uses the feedback from an expert user to efficiently search for relevant anomalies. Our results, using synthetic and real datasets, indicate that, under noisy data and anomalies presenting regular patterns, our approach correctly identifies relevant anomalies. \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Mixing Hierarchical Contexts for Object Recognition.\n \n\n\n \n Peralta, B.; and Soto, A.\n\n\n \n\n\n\n In CIARP, 2011. \n \n\n\n\n
\n\n\n\n \n \n \"MixingPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 2 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Peralta:EtAl:2011,\n  Author = {B. Peralta and A. Soto},\n  Title = {Mixing Hierarchical Contexts for Object Recognition},\n  booktitle = {{CIARP}},\n  year = {2011},\n    abstract = {Robust category-level object recognition is currently a major\ngoal for the Computer Vision community. Intra-class and pose variations,\nas well as, background clutter and partial occlusions are some of the main\ndifficulties to achieve this goal. Contextual information in the form of ob-\nject co-ocurrences and spatial contraints has been successfully applied to\nreduce the inherent uncertainty of the visual world. Recently, Choi et\nal. [5] propose the use of a tree-structured graphical model to capture\ncontextual relations among objects. Under this model there is only one\npossible fixed contextual relation among subsets of objects. In this work\nwe extent Choi et al. approach by using a mixture model to consider the\ncase that contextual relations among objects depend on scene type. Our\nexperiments highlight the advantages of our proposal, showing that the\nadaptive specialization of contextual relations improves object recogni-\ntion and object detection performances.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/Peralta-2011.pdf}\n}\n\n\n\n%***********2010***************%\n\n\n
\n
\n\n\n
\n Robust category-level object recognition is currently a major goal for the Computer Vision community. Intra-class and pose variations, as well as, background clutter and partial occlusions are some of the main difficulties to achieve this goal. Contextual information in the form of ob- ject co-ocurrences and spatial contraints has been successfully applied to reduce the inherent uncertainty of the visual world. Recently, Choi et al. [5] propose the use of a tree-structured graphical model to capture contextual relations among objects. Under this model there is only one possible fixed contextual relation among subsets of objects. In this work we extent Choi et al. approach by using a mixture model to consider the case that contextual relations among objects depend on scene type. Our experiments highlight the advantages of our proposal, showing that the adaptive specialization of contextual relations improves object recogni- tion and object detection performances. \n
\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 2010\n \n \n (6)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n Face Recognition with Decision Tree-based Local Binary Patterns.\n \n\n\n \n Maturana, D.; Mery, D.; and Soto, A.\n\n\n \n\n\n\n In Proc. of Asian Conference on Computer Vision (ACCV-2010), 2010. \n \n\n\n\n
\n\n\n\n \n \n \"FacePaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 2 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Maturana:EtAl:2010,\n  Author = {D. Maturana and D. Mery and A. Soto},\n  Title = {Face Recognition with Decision Tree-based Local Binary Patterns},\n  booktitle = {Proc. of Asian Conference on Computer Vision (ACCV-2010)},\n  year = {2010},\n    abstract = {Many state-of-the-art face recognition algorithms use image\ndescriptors based on features known as Local Binary Patterns (LBPs).\nWhile many variations of LBP exist, so far none of them can automati-\ncally adapt to the training data. We introduce and analyze a novel gen-\neralization of LBP that learns the most discriminative LBP-like features\nfor each facial region in a supervised manner. Since the proposed method\nis based on Decision Trees, we call it Decision Tree Local Binary Pat-\nterns or DT-LBPs. Tests on standard face recognition datasets show the\nsuperiority of DT-LBP with respect of several state-of-the-art feature\ndescriptors regularly used in face recognition applications.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/ACCV-2010.pdf}\n}\n\n
\n
\n\n\n
\n Many state-of-the-art face recognition algorithms use image descriptors based on features known as Local Binary Patterns (LBPs). While many variations of LBP exist, so far none of them can automati- cally adapt to the training data. We introduce and analyze a novel gen- eralization of LBP that learns the most discriminative LBP-like features for each facial region in a supervised manner. Since the proposed method is based on Decision Trees, we call it Decision Tree Local Binary Pat- terns or DT-LBPs. Tests on standard face recognition datasets show the superiority of DT-LBP with respect of several state-of-the-art feature descriptors regularly used in face recognition applications. \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Automated Detection of Fish Bones in Salmon Fillets using X-ray Testing.\n \n\n\n \n Mery, D.; Lillo, I.; Loebel, H.; Riffo, V.; Soto, A.; Cipriano, A.; and Aguilera, J.\n\n\n \n\n\n\n In Proc. of 4th Pacific-Rim Symposium on Image and Video Technology (PSIVT-2010), 2010. \n \n\n\n\n
\n\n\n\n \n \n \"AutomatedPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 1 download\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Mery-PSIVT:EtAl:2010,\n  Author = {D. Mery and I. Lillo and H. Loebel and V. Riffo and A. Soto and A.\nCipriano and JM. Aguilera},\n  Title = {Automated Detection of Fish Bones in Salmon Fillets using X-ray\nTesting},\n  booktitle = {Proc. of 4th Pacific-Rim Symposium on Image and Video Technology\n(PSIVT-2010)},\n  year = {2010},\n    abstract = {X-ray testing is playing an increasingly important\nrole in food quality assurance. In the production of fish fillets,\nhowever, fish bone detection is performed by human operators\nusing their sense of touch and vision which can lead to\nmisclassification. In countries where fish is often consumed, fish\nbones are some of the most frequently ingested foreign bodies\nencountered in foods. Effective detection of fish bones in the\nquality control process would help avoid this problem. For this\nreason, we developed an X-ray machine vision approach to au-\ntomatically detect fish bones in fish fillets. This paper describes\nour approach and the corresponding validation experiments\nwith salmon fillets. The approach consists of six steps: 1) A\ndigital X-ray image is taken of the fish fillet being tested. 2) The\nX-ray image is filtered and enhanced to facilitate the detection\nof fish bones. 3) Potential fish bones in the image are segmented\nusing band pass filtering, thresholding and morphological\ntechniques. 4) Intensity features of the enhanced X-ray image\nare extracted from small detection windows that are defined\nin those regions where potential fish bones were segmented.\n5) A classifier is used to discriminate between ‘bones’ and\n‘no-bones’ classes in the detection windows. 6) Finally, fish\nbones in the X-ray image are isolated using morphological\noperations applied on the corresponding segments classified as\n‘bones’. In the experiments we used a high resolution flat panel\ndetector with the capacity to capture up to a 6 million pixel\ndigital X-ray image. In the training phase, we analyzed 20\nrepresentative salmon fillets, 7700 detection windows (10×10\npixels) and 279 intensity features. Cross validation yielded a\ndetection performance of 95% using a support vector machine\nclassifier with only 24 selected features. We believe that the\nproposed approach opens new possibilities in the field of\nautomated visual inspection of salmon and other similar fish.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/PSIVT-2010.pdf}\n}\n\n
\n
\n\n\n
\n X-ray testing is playing an increasingly important role in food quality assurance. In the production of fish fillets, however, fish bone detection is performed by human operators using their sense of touch and vision which can lead to misclassification. In countries where fish is often consumed, fish bones are some of the most frequently ingested foreign bodies encountered in foods. Effective detection of fish bones in the quality control process would help avoid this problem. For this reason, we developed an X-ray machine vision approach to au- tomatically detect fish bones in fish fillets. This paper describes our approach and the corresponding validation experiments with salmon fillets. The approach consists of six steps: 1) A digital X-ray image is taken of the fish fillet being tested. 2) The X-ray image is filtered and enhanced to facilitate the detection of fish bones. 3) Potential fish bones in the image are segmented using band pass filtering, thresholding and morphological techniques. 4) Intensity features of the enhanced X-ray image are extracted from small detection windows that are defined in those regions where potential fish bones were segmented. 5) A classifier is used to discriminate between ‘bones’ and ‘no-bones’ classes in the detection windows. 6) Finally, fish bones in the X-ray image are isolated using morphological operations applied on the corresponding segments classified as ‘bones’. In the experiments we used a high resolution flat panel detector with the capacity to capture up to a 6 million pixel digital X-ray image. In the training phase, we analyzed 20 representative salmon fillets, 7700 detection windows (10×10 pixels) and 279 intensity features. Cross validation yielded a detection performance of 95% using a support vector machine classifier with only 24 selected features. We believe that the proposed approach opens new possibilities in the field of automated visual inspection of salmon and other similar fish. \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Quality Classification of Corn Tortillas using Computer Vision.\n \n\n\n \n Mery, D.; Chanona-Perez, J.; Soto, A.; Aguilera, J.; Cipriano, A.; Velez-Riverab, N.; Arzate-Vazquez, I.; and Gutierrez-Lopez, G.\n\n\n \n\n\n\n Journal of Food Engineering, 101(4): 357-364. 2010.\n \n\n\n\n
\n\n\n\n \n \n \"QualityPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 1 download\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Mery:EtAl:2010,\n  Author = {D. Mery and J. Chanona-Perez and A. Soto and JM. Aguilera and A.\nCipriano and N. Velez-Riverab and I. Arzate-Vazquez and G. Gutierrez-Lopez},\n  Title = {Quality Classification of Corn Tortillas using Computer Vision},\n  Journal = {Journal of Food Engineering},\n  Volume = {101},\n  Number = {4},\n  Pages = {357-364},\n  Year = {2010},\n    abstract = {Computer vision is playing an increasingly important role in automated visual food inspection. However quality control in tortilla\nproduction is still performed by human operators which may lead to misclassification due to their subjectivity and fatigue. In order to\nreduce the need for human operators and therefore misclassification, we developed a computer vision framework to automatically\nclassify the quality of corn tortillas according to five hedonic sub-classes given by a sensorial panel. The proposed framework\nanalyzed 750 corn tortillas obtained from 15 different Mexican commercial stores which were either small, medium or large in\nsize. More than 2300 geometric and color features were extracted from 1500 images capturing both sides of the 750 tortillas.\nAfter implementing a feature selection algorithm, in which the most relevant features were selected for the classification of the\nfive sub-classes, only 64 features were required to design a classifier based on support vector machines. Cross validation yielded a\nperformance of 95% in the classification of the five hedonic sub-classes. Additionally, using only 10 of the selected features and a\nsimple statistical classifier, it was possible to determine the origin of the tortillas with a performance of 96%. We believe that the\nproposed framework opens up new possibilities in the field of automated visual inspection of tortillas.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/Tortillas-2010.pdf}\n}\n\n
\n
\n\n\n
\n Computer vision is playing an increasingly important role in automated visual food inspection. However quality control in tortilla production is still performed by human operators which may lead to misclassification due to their subjectivity and fatigue. In order to reduce the need for human operators and therefore misclassification, we developed a computer vision framework to automatically classify the quality of corn tortillas according to five hedonic sub-classes given by a sensorial panel. The proposed framework analyzed 750 corn tortillas obtained from 15 different Mexican commercial stores which were either small, medium or large in size. More than 2300 geometric and color features were extracted from 1500 images capturing both sides of the 750 tortillas. After implementing a feature selection algorithm, in which the most relevant features were selected for the classification of the five sub-classes, only 64 features were required to design a classifier based on support vector machines. Cross validation yielded a performance of 95% in the classification of the five hedonic sub-classes. Additionally, using only 10 of the selected features and a simple statistical classifier, it was possible to determine the origin of the tortillas with a performance of 96%. We believe that the proposed framework opens up new possibilities in the field of automated visual inspection of tortillas. \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Active visual perception for mobile robot localization.\n \n\n\n \n Correa, J.; and Soto, A.\n\n\n \n\n\n\n Journal of Intelligent and Robotic Systems, 58(3-4): 339-354. 2010.\n \n\n\n\n
\n\n\n\n \n \n \"ActivePaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 1 download\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Correa:EtAl:2010,\n  Author = {J. Correa and A. Soto},\n  Title = {Active visual perception for mobile robot localization},\n  Journal = {Journal of Intelligent and Robotic Systems},\n  Volume = {58},\n  Number = {3-4},\n  Pages = {339-354},\n  Year = {2010},\n    abstract = {Localization is a key issue for a mobile robot, in particular in environments\nwhere a globally accurate positioning system, such as GPS, is not available. In these\nenvironments, accurate and efficient robot localization is not a trivial task, as an\nincrease in accuracy usually leads to an impoverishment in efficiency and viceversa.\nActive perception appears as an appealing way to improve the localization process\nby increasing the richness of the information acquired from the environment. In this\npaper, we present an active perception strategy for a mobile robot provided with\na visual sensor mounted on a pan-tilt mechanism. The visual sensor has a limited\nfield of view, so the goal of the active perception strategy is to use the pan-tilt\nunit to direct the sensor to informative parts of the environment. To achieve this\ngoal, we use a topological map of the environment and a Bayesian non-parametric\nestimation of robot position based on a particle filter. We slightly modify the regular\nimplementation of this filter by including an additional step that selects the best\nperceptual action using Monte Carlo estimations. We understand the best perceptual\naction as the one that produces the greatest reduction in uncertainty about the\nrobot position. We also consider in our optimization function a cost term that\nfavors efficient perceptual actions. Previous works have proposed active perception\nstrategies for robot localization, but mainly in the context of range sensors, grid\nrepresentations of the environment, and parametric techniques, such as the extended\nKalman filter. Accordingly, the main contributions of this work are: i) Development\nof a sound strategy for active selection of perceptual actions in the context of a visual\nsensor and a topological map; ii) Real time operation using a modified version of the\nparticle filter and Monte Carlo based estimations; iii) Implementation and testing\nof these ideas using simulations and a real case scenario. Our results indicate that,\nin terms of accuracy of robot localization, the proposed approach decreases mean\naverage error and standard deviation with respect to a passive perception scheme.\nFurthermore, in terms of efficiency, the active scheme is able to operate in real time\nwithout adding a relevant overhead to the regular robot operation.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/Intell-Robots-2010.pdf}\n}\n\n\n
\n
\n\n\n
\n Localization is a key issue for a mobile robot, in particular in environments where a globally accurate positioning system, such as GPS, is not available. In these environments, accurate and efficient robot localization is not a trivial task, as an increase in accuracy usually leads to an impoverishment in efficiency and viceversa. Active perception appears as an appealing way to improve the localization process by increasing the richness of the information acquired from the environment. In this paper, we present an active perception strategy for a mobile robot provided with a visual sensor mounted on a pan-tilt mechanism. The visual sensor has a limited field of view, so the goal of the active perception strategy is to use the pan-tilt unit to direct the sensor to informative parts of the environment. To achieve this goal, we use a topological map of the environment and a Bayesian non-parametric estimation of robot position based on a particle filter. We slightly modify the regular implementation of this filter by including an additional step that selects the best perceptual action using Monte Carlo estimations. We understand the best perceptual action as the one that produces the greatest reduction in uncertainty about the robot position. We also consider in our optimization function a cost term that favors efficient perceptual actions. Previous works have proposed active perception strategies for robot localization, but mainly in the context of range sensors, grid representations of the environment, and parametric techniques, such as the extended Kalman filter. Accordingly, the main contributions of this work are: i) Development of a sound strategy for active selection of perceptual actions in the context of a visual sensor and a topological map; ii) Real time operation using a modified version of the particle filter and Monte Carlo based estimations; iii) Implementation and testing of these ideas using simulations and a real case scenario. Our results indicate that, in terms of accuracy of robot localization, the proposed approach decreases mean average error and standard deviation with respect to a passive perception scheme. Furthermore, in terms of efficiency, the active scheme is able to operate in real time without adding a relevant overhead to the regular robot operation. \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Indoor Scene Recognition Through Object Detection.\n \n\n\n \n Espinace, P.; Kollar, T.; Soto, A.; and Roy, N.\n\n\n \n\n\n\n In Proc. of IEEE Int. Conf. on Robotics and Automation (ICRA), 2010. \n \n\n\n\n
\n\n\n\n \n \n \"IndoorPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 2 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Espinace:EtAl:2010,\n  Author = {P. Espinace and T. Kollar and A. Soto and N. Roy},\n  Title = {Indoor Scene Recognition Through Object Detection},\n  booktitle = {Proc. of IEEE Int. Conf. on Robotics and Automation (ICRA)},\n  year = {2010},\n    abstract = {Scene recognition is a highly valuable percep-\ntual ability for an indoor mobile robot, however, current\napproaches for scene recognition present a significant drop in\nperformance for the case of indoor scenes. We believe that\nthis can be explained by the high appearance variability of\nindoor environments. This stresses the need to include high-\nlevel semantic information in the recognition process. In this\nwork we propose a new approach for indoor scene recognition\nbased on a generative probabilistic hierarchical model that uses\ncommon objects as an intermediate semantic representation.\nUnder this model, we use object classifiers to associate low-\nlevel visual features to objects, and at the same time, we use\ncontextual relations to associate objects to scenes. As a further\ncontribution, we improve the performance of current state-of-\nthe-art category-level object classifiers by including geometrical\ninformation obtained from a 3D range sensor that facilitates\nthe implementation of a focus of attention mechanism within a\nMonte Carlo sampling scheme. We test our approach using real\ndata, showing significant advantages with respect to previous\nstate-of-the-art methods.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/Icra-2010.pdf}\n}\n\n
\n
\n\n\n
\n Scene recognition is a highly valuable percep- tual ability for an indoor mobile robot, however, current approaches for scene recognition present a significant drop in performance for the case of indoor scenes. We believe that this can be explained by the high appearance variability of indoor environments. This stresses the need to include high- level semantic information in the recognition process. In this work we propose a new approach for indoor scene recognition based on a generative probabilistic hierarchical model that uses common objects as an intermediate semantic representation. Under this model, we use object classifiers to associate low- level visual features to objects, and at the same time, we use contextual relations to associate objects to scenes. As a further contribution, we improve the performance of current state-of- the-art category-level object classifiers by including geometrical information obtained from a 3D range sensor that facilitates the implementation of a focus of attention mechanism within a Monte Carlo sampling scheme. We test our approach using real data, showing significant advantages with respect to previous state-of-the-art methods. \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Human Detection Using a Mobile Platform and Novel Features Derived From a Visual Saliency Mechanism.\n \n\n\n \n Montabone, S.; and Soto, A.\n\n\n \n\n\n\n Image and Vision Computing, 28(3): 391-402. 2010.\n \n\n\n\n
\n\n\n\n \n \n \"HumanPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 4 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Montabone:EtAl:2010,\n  Author = {S. Montabone and A. Soto},\n  Title = {Human Detection Using a Mobile Platform and Novel Features Derived\nFrom a Visual Saliency Mechanism},\n  Journal = {Image and Vision Computing},\n  Volume = {28},\n  Number = {3},\n  Pages = {391-402},\n  Year = {2010},\n    abstract = {Human detection is a key ability to an increasing number of applications that operates in human inhab-\nited environments or needs to interact with a human user. Currently, most successful approaches to\nhuman detection are based on background substraction techniques that apply only to the case of static\ncameras or cameras with highly constrained motions. Furthermore, many applications rely on features\nderived from specific human poses, such as systems based on features derived from the human face\nwhich is only visible when a person is facing the detecting camera. In this work, we present a new com-\nputer vision algorithm designed to operate with moving cameras and to detect humans in different poses\nunder partial or complete view of the human body. We follow a standard pattern recognition approach\nbased on four main steps: (i) preprocessing to achieve color constancy and stereo pair calibration, (ii) seg-\nmentation using depth continuity information, (iii) feature extraction based on visual saliency, and (iv)\nclassification using a neural network. The main novelty of our approach lies in the feature extraction step,\nwhere we propose novel features derived from a visual saliency mechanism. In contrast to previous\nworks, we do not use a pyramidal decomposition to run the saliency algorithm, but we implement this\nat the original image resolution using the so-called integral image. Our results indicate that our method:\n(i) outperforms state-of-the-art techniques for human detection based on face detectors, (ii) outperforms\nstate-of-the-art techniques for complete human body detection based on different set of visual features,\nand (iii) operates in real time onboard a mobile platform, such as a mobile robot (15 fps).\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/ImageVisionComp-10.pdf}\n}\n\n\n%***********2009 and older***************%\n\n
\n
\n\n\n
\n Human detection is a key ability to an increasing number of applications that operates in human inhab- ited environments or needs to interact with a human user. Currently, most successful approaches to human detection are based on background substraction techniques that apply only to the case of static cameras or cameras with highly constrained motions. Furthermore, many applications rely on features derived from specific human poses, such as systems based on features derived from the human face which is only visible when a person is facing the detecting camera. In this work, we present a new com- puter vision algorithm designed to operate with moving cameras and to detect humans in different poses under partial or complete view of the human body. We follow a standard pattern recognition approach based on four main steps: (i) preprocessing to achieve color constancy and stereo pair calibration, (ii) seg- mentation using depth continuity information, (iii) feature extraction based on visual saliency, and (iv) classification using a neural network. The main novelty of our approach lies in the feature extraction step, where we propose novel features derived from a visual saliency mechanism. In contrast to previous works, we do not use a pyramidal decomposition to run the saliency algorithm, but we implement this at the original image resolution using the so-called integral image. Our results indicate that our method: (i) outperforms state-of-the-art techniques for human detection based on face detectors, (ii) outperforms state-of-the-art techniques for complete human body detection based on different set of visual features, and (iii) operates in real time onboard a mobile platform, such as a mobile robot (15 fps). \n
\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 2009\n \n \n (4)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n Collaborative Robotic Instruction: A Graph Teaching Experience.\n \n\n\n \n Mitnik, R.; Recabarren, M.; Nussbaum, M.; and Soto, A.\n\n\n \n\n\n\n Computers and Education, 53(2): 330-342. 2009.\n \n\n\n\n
\n\n\n\n \n \n \"CollaborativePaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 3 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Mitnik:EtAl:2009,\n  Author = {R. Mitnik and M. Recabarren and M. Nussbaum and A. Soto},\n  Title = {Collaborative Robotic Instruction: A Graph Teaching Experience},\n  Journal = {Computers and Education},\n  Volume = {53},\n  Number = {2},\n  Pages = {330-342},\n  Year = {2009},\n  abstract = {Graphing is a key skill in the study of Physics. Drawing and interpreting graphs play a key role in the\nunderstanding of science, while the lack of these has proved to be a handicap and a limiting factor in\nthe learning of scientific concepts. It has been observed that despite the amount of previous graph-work-\ning experience, students of all ages experience a series of difficulties when trying to comprehend graphs\nor when trying to relate them with physical concepts such as position, velocity and acceleration. Several\ncomputational tools have risen to improve the students’ understanding of kinematical graphs; however,\nthese approaches fail to develop graph construction skills. On the other hand, Robots have opened new\nopportunities in learning. Nevertheless, most of their educational applications focus on Robotics related\nsubjects, such as robot programming, robot construction, and artificial intelligence. This paper describes a\nrobotic activity based on face-to-face computer supported collaborative learning. By means of a set of\nhandhelds and a robot wirelessly interconnected, the aim of the activity is to develop graph construction\nand graph interpretation skills while also reinforcing kinematics concepts. Results show that students\nusing the robotic activity achieve a significant increase in their graph interpreting skills. Moreover, when\ncompared with a similar computer-simulated activity, it proved to be almost twice as effective. Finally,\nthe robotic application proved to be a highly motivating activity for the students, fostering collaboration\namong them.},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/Mitnik_2009_Computers-&-Education.pdf}\n}\n\n
\n
\n\n\n
\n Graphing is a key skill in the study of Physics. Drawing and interpreting graphs play a key role in the understanding of science, while the lack of these has proved to be a handicap and a limiting factor in the learning of scientific concepts. It has been observed that despite the amount of previous graph-work- ing experience, students of all ages experience a series of difficulties when trying to comprehend graphs or when trying to relate them with physical concepts such as position, velocity and acceleration. Several computational tools have risen to improve the students’ understanding of kinematical graphs; however, these approaches fail to develop graph construction skills. On the other hand, Robots have opened new opportunities in learning. Nevertheless, most of their educational applications focus on Robotics related subjects, such as robot programming, robot construction, and artificial intelligence. This paper describes a robotic activity based on face-to-face computer supported collaborative learning. By means of a set of handhelds and a robot wirelessly interconnected, the aim of the activity is to develop graph construction and graph interpretation skills while also reinforcing kinematics concepts. Results show that students using the robotic activity achieve a significant increase in their graph interpreting skills. Moreover, when compared with a similar computer-simulated activity, it proved to be almost twice as effective. Finally, the robotic application proved to be a highly motivating activity for the students, fostering collaboration among them.\n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Performance Evaluation of the Covariance Descriptor for Target Detection.\n \n\n\n \n Cortez-Cargill, P.; Undurraga-Rius, C.; Mery, D.; and Soto, A.\n\n\n \n\n\n\n In Proc. of XXVIII Int. Conf. of the Chilean Computer Science Society/IEEE CS Press, 2009. \n \n\n\n\n
\n\n\n\n \n \n \"PerformancePaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 1 download\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Cortez:EtAl:2009,\n  Author = {P. Cortez-Cargill and C. Undurraga-Rius and D. Mery and A. Soto},\n  Title = {Performance Evaluation of the Covariance Descriptor for Target\nDetection},\n  booktitle = {Proc. of XXVIII Int. Conf. of the Chilean Computer Science\nSociety/IEEE CS Press},\n  year = {2009},\n    abstract = {In computer vision, there has been a strong advance\nin creating new image descriptors. A descriptor that has recently\nappeared is the Covariance Descriptor, but there have not been\nany studies about the different methodologies for its construction.\nTo address this problem we have conducted an analysis on the\ncontribution of diverse features of an image to the descriptor and\ntherefore their contribution to the detection of varied targets, in\nour case: faces and pedestrians. That is why we have defined a\nmethodology to determinate the performance of the covariance\nmatrix created from different characteristics. Now we are able to\ndeterminate the best set of features for face and people detection,\nfor each problem. We have also achieved to establish that not\nany kind of combination of features can be used because it might\nnot exist a correlation between them. Finally, when an analysis\nis performed with the best set of features, for the face detection\nproblem we reach a performance of 99%, meanwhile for the\npedestrian detection problem we reach a performance of 85%.\nWith this we hope we have built a more solid base when choosing\nfeatures for this descriptor, allowing to move forward to other\ntopics such as object recognition or tracking.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/Final-Proceedings-Cortez_Undurraga_Mery_Soto_SCCC2009.pdf}\n}\n\n
\n
\n\n\n
\n In computer vision, there has been a strong advance in creating new image descriptors. A descriptor that has recently appeared is the Covariance Descriptor, but there have not been any studies about the different methodologies for its construction. To address this problem we have conducted an analysis on the contribution of diverse features of an image to the descriptor and therefore their contribution to the detection of varied targets, in our case: faces and pedestrians. That is why we have defined a methodology to determinate the performance of the covariance matrix created from different characteristics. Now we are able to determinate the best set of features for face and people detection, for each problem. We have also achieved to establish that not any kind of combination of features can be used because it might not exist a correlation between them. Finally, when an analysis is performed with the best set of features, for the face detection problem we reach a performance of 99%, meanwhile for the pedestrian detection problem we reach a performance of 85%. With this we hope we have built a more solid base when choosing features for this descriptor, allowing to move forward to other topics such as object recognition or tracking. \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n An ensemble of Discriminative Local Subspaces in Microarray Data for Gene Ontology Annotation Predictions.\n \n\n\n \n Puelma, T.; Soto, A.; and Gutierrez, R.\n\n\n \n\n\n\n In Proc. of 1st Chilean Workshop on Pattern Recognition (CWPR), pages 52-61, 2009. \n \n\n\n\n
\n\n\n\n \n \n \"AnPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 1 download\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Puelma:EtAl:2009,\n  Author = {T. Puelma and A. Soto and R. Gutierrez},\n  Title = {An ensemble of Discriminative Local Subspaces in Microarray Data for\nGene Ontology Annotation Predictions},\n  booktitle = {Proc. of 1st Chilean Workshop on Pattern Recognition (CWPR)},\n  pages = {52-61},\n  year = {2009},\n    abstract = {Genome sequencing has allowed to know almost\nevery gene of many organisms. However, understanding the\nfunctions of most genes is still an open problem. In this\npaper, we present a novel machine learning method to predict\nfunctions of unknown genes in base of gene expression data\nand Gene Ontology annotations. Most function prediction al-\ngorithms developed in the past don’t exploit the discriminative\npower of supervised learning. In contrast, our method uses\nthis to find discriminative local subspaces that are suitable to\nperform gene functional prediction. Cross-validation test are\ndone in artificial and real data and compared with a state-of-\nthe-art method. Preliminary results shows that in overall, our\nmethod outperforms the other approach in terms of precision\nand recall, giving insights in the importance of a good selection\nof discriminative experiments.},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/DLS-Final-v2.pdf}\n}\n\n
\n
\n\n\n
\n Genome sequencing has allowed to know almost every gene of many organisms. However, understanding the functions of most genes is still an open problem. In this paper, we present a novel machine learning method to predict functions of unknown genes in base of gene expression data and Gene Ontology annotations. Most function prediction al- gorithms developed in the past don’t exploit the discriminative power of supervised learning. In contrast, our method uses this to find discriminative local subspaces that are suitable to perform gene functional prediction. Cross-validation test are done in artificial and real data and compared with a state-of- the-art method. Preliminary results shows that in overall, our method outperforms the other approach in terms of precision and recall, giving insights in the importance of a good selection of discriminative experiments.\n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Face Recognition with Local Binary Patterns, Spatial Pyramid Histograms and Naive Bayes Nearest Neighbor classification.\n \n\n\n \n Maturana, D.; Mery, D.; and Soto, A.\n\n\n \n\n\n\n In Proc. of XXVIII Int. Conf. of the Chilean Computer Science Society/IEEE CS Press, 2009. \n \n\n\n\n
\n\n\n\n \n \n \"FacePaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 3 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Maturana:EtAl:2009,\n  Author = {D. Maturana and D. Mery and A. Soto},\n  Title = {Face Recognition with Local Binary Patterns, Spatial Pyramid\nHistograms and Naive Bayes Nearest Neighbor classification},\n  booktitle = {Proc. of XXVIII Int. Conf. of the Chilean Computer Science\nSociety/IEEE CS Press},\n  year = {2009},\n    abstract = {Face recognition algorithms commonly assume\nthat face images are well aligned and have a similar pose\n– yet in many practical applications it is impossible to meet\n  these conditions. Therefore extending face recognition to un-\n  constrained face images has become an active area of research.\n  To this end, histograms of Local Binary Patterns (LBP)\n  have proven to be highly discriminative descriptors for face\n  recognition. Nonetheless, most LBP-based algorithms use a\n  rigid descriptor matching strategy that is not robust against\n  pose variation and misalignment.\n  We propose two algorithms for face recognition that are de-\n  signed to deal with pose variations and misalignment. We also\n  incorporate an illumination normalization step that increases\n  robustness against lighting variations. The proposed algorithms\n  use descriptors based on histograms of LBP and perform\n  descriptor matching with spatial pyramid matching (SPM) and\n  Naive Bayes Nearest Neighbor (NBNN), respectively. Our con-\n  tribution is the inclusion of flexible spatial matching schemes\n    that use an image-to-class relation to provide an improved\n    robustness with respect to intra-class variations.\n    We compare the accuracy of the proposed algorithms against\n    Ahonen’s original LBP-based face recognition system and two\n      baseline holistic classifiers on four standard datasets. Our\n        results indicate that the algorithm based on NBNN outperforms\n        the other solutions, and does so more markedly in presence of\n        pose variations.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/Final-Daniel-09.pdf}\n}\n\n\n\n%***********2009 and older***************%\n\n\n
\n
\n\n\n
\n Face recognition algorithms commonly assume that face images are well aligned and have a similar pose – yet in many practical applications it is impossible to meet these conditions. Therefore extending face recognition to un- constrained face images has become an active area of research. To this end, histograms of Local Binary Patterns (LBP) have proven to be highly discriminative descriptors for face recognition. Nonetheless, most LBP-based algorithms use a rigid descriptor matching strategy that is not robust against pose variation and misalignment. We propose two algorithms for face recognition that are de- signed to deal with pose variations and misalignment. We also incorporate an illumination normalization step that increases robustness against lighting variations. The proposed algorithms use descriptors based on histograms of LBP and perform descriptor matching with spatial pyramid matching (SPM) and Naive Bayes Nearest Neighbor (NBNN), respectively. Our con- tribution is the inclusion of flexible spatial matching schemes that use an image-to-class relation to provide an improved robustness with respect to intra-class variations. We compare the accuracy of the proposed algorithms against Ahonen’s original LBP-based face recognition system and two baseline holistic classifiers on four standard datasets. Our results indicate that the algorithm based on NBNN outperforms the other solutions, and does so more markedly in presence of pose variations. \n
\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 2008\n \n \n (7)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n Detection of Anomalies in Large Datasets Using an Active Learning Scheme Based on Dirichlet Distributions.\n \n\n\n \n Pichara, K.; Soto, A.; and Araneda, A.\n\n\n \n\n\n\n In Advances in Artificial Intelligence, Iberamia-08, LNCS 5290, pages 163-172, 2008. \n \n\n\n\n
\n\n\n\n \n \n \"DetectionPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 1 download\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Pichara:EtAl:2008,\n  Author = {K. Pichara and A. Soto and A. Araneda},\n  Title = {Detection of Anomalies in Large Datasets Using an Active Learning\nScheme Based on Dirichlet Distributions},\n  booktitle = {Advances in Artificial Intelligence, Iberamia-08, LNCS 5290},\n  pages = {163-172},\n  year = {2008},\n    abstract = {Today, the detection of anomalous records is a highly valu-\nable application in the analysis of current huge datasets. In this paper\nwe propose a new algorithm that, with the help of a human expert, effi-\nciently explores a dataset with the goal of detecting relevant anomalous\nrecords. Under this scheme the computer selectively asks the expert for\ndata labeling, looking for relevant semantic feedback in order to improve\nits knowledge about what characterizes a relevant anomaly. Our ratio-\nnale is that while computers can process huge amounts of low level data,\nan expert has high level semantic knowledge to efficiently lead the search.\nWe build upon our previous work based on Bayesian networks that pro-\nvides an initial set of potential anomalies. In this paper, we augment this\napproach with an active learning scheme based on the clustering proper-\nties of Dirichlet distributions. We test the performance of our algorithm\nusing synthetic and real datasets. Our results indicate that, under noisy\ndata and anomalies presenting regular patterns, our approach signifi-\ncantly reduces the rate of false positives, while decreasing the time to\nreach the relevant anomalies.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/ActiveLearning.pdf}\n}\n\n
\n
\n\n\n
\n Today, the detection of anomalous records is a highly valu- able application in the analysis of current huge datasets. In this paper we propose a new algorithm that, with the help of a human expert, effi- ciently explores a dataset with the goal of detecting relevant anomalous records. Under this scheme the computer selectively asks the expert for data labeling, looking for relevant semantic feedback in order to improve its knowledge about what characterizes a relevant anomaly. Our ratio- nale is that while computers can process huge amounts of low level data, an expert has high level semantic knowledge to efficiently lead the search. We build upon our previous work based on Bayesian networks that pro- vides an initial set of potential anomalies. In this paper, we augment this approach with an active learning scheme based on the clustering proper- ties of Dirichlet distributions. We test the performance of our algorithm using synthetic and real datasets. Our results indicate that, under noisy data and anomalies presenting regular patterns, our approach signifi- cantly reduces the rate of false positives, while decreasing the time to reach the relevant anomalies. \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n An autonomous educational mobile robot mediator.\n \n\n\n \n Mitnik, R.; Nussbaum, M.; and Soto, A.\n\n\n \n\n\n\n Autonomous Robots, 25(4): 367-382. 2008.\n \n\n\n\n
\n\n\n\n \n \n \"AnPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 1 download\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Mitnik:EtAl:2008,\n  Author = {R. Mitnik and M. Nussbaum and A. Soto},\n  Title = {An autonomous educational mobile robot mediator},\n  Journal = {Autonomous Robots},\n  Volume = {25},\n  Number = {4},\n  Pages = {367-382},\n  Year = {2008},\n    abstract = {So far, most of the applications of robotic\ntechnology to education have mainly focused on sup-\nporting the teaching of subjects that are closely related\nto the Robotics field, such as robot programming, robot\nconstruction, or mechatronics. Moreover, most of the\napplications have used the robot as an end or a passive\ntool of the learning activity, where the robot has been\nconstructed or programmed. In this paper, we present a\nnovel application of robotic technologies to education,\nwhere we use the real world situatedness of a robot to\nteach non-robotic related subjects, such as math and\nphysics. Furthermore, we also provide the robot with\na suitable degree of autonomy to actively guide and\nmediate in the development of the educational activ-\nity. We present our approach as an educational frame-\nwork based on a collaborative and constructivist learn-\ning environment, where the robot is able to act as an\ninteraction mediator capable of managing the interac-\ntions occurring among the working students. We illus-\ntrate the use of this framework by a 4-step methodology\nthat is used to implement two educational activities.\nThese activities were tested at local schools with en-\ncouraging results. Accordingly, the main contributions\nof this work are: i) A novel use of a mobile robot to\nillustrate and teach relevant concepts and properties of\nthe real world; ii) A novel use of robots as mediators\nthat autonomously guide an educational activity using\na collaborative and constructivist learning approach;\niii) The implementation and testing of these ideas in\na real scenario, working with students at local schools.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/EducationalRobot.pdf}\n}\n\n
\n
\n\n\n
\n So far, most of the applications of robotic technology to education have mainly focused on sup- porting the teaching of subjects that are closely related to the Robotics field, such as robot programming, robot construction, or mechatronics. Moreover, most of the applications have used the robot as an end or a passive tool of the learning activity, where the robot has been constructed or programmed. In this paper, we present a novel application of robotic technologies to education, where we use the real world situatedness of a robot to teach non-robotic related subjects, such as math and physics. Furthermore, we also provide the robot with a suitable degree of autonomy to actively guide and mediate in the development of the educational activ- ity. We present our approach as an educational frame- work based on a collaborative and constructivist learn- ing environment, where the robot is able to act as an interaction mediator capable of managing the interac- tions occurring among the working students. We illus- trate the use of this framework by a 4-step methodology that is used to implement two educational activities. These activities were tested at local schools with en- couraging results. Accordingly, the main contributions of this work are: i) A novel use of a mobile robot to illustrate and teach relevant concepts and properties of the real world; ii) A novel use of robots as mediators that autonomously guide an educational activity using a collaborative and constructivist learning approach; iii) The implementation and testing of these ideas in a real scenario, working with students at local schools. \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Unsupervised Identification of Useful Visual Landmarks Using Multiple Segmentations and Top-Down Feedback.\n \n\n\n \n Espinace, P.; Langdon, D.; and Soto, A.\n\n\n \n\n\n\n Robotics and Autonomous Systems, 56(6): 538-548. 2008.\n \n\n\n\n
\n\n\n\n \n \n \"UnsupervisedPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 1 download\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Espinace:Soto:2008a,\n  Author = {P. Espinace and D. Langdon and A. Soto},\n  Title = {Unsupervised Identification of Useful Visual Landmarks Using Multiple\nSegmentations and Top-Down Feedback},\n  Journal = {Robotics and Autonomous Systems},\n  Volume = {56},\n  Number = {6},\n  Pages = {538-548},\n  Year = {2008},\n    abstract = {In this paper, we tackle the problem of unsupervised selection and posterior recognition of visual landmarks in images sequences\nacquired by an indoor mobile robot. This is a highly valuable perceptual capability for a wide variety of robotic applications, in\nparticular autonomous navigation. Our method combines a bottom-up data driven approach with top-down feedback provided\nby high level semantic representations. The bottom-up approach is based on three main mechanisms: visual attention, area\nsegmentation, and landmark characterization. As there is no segmentation method that works properly in every situation, we\nintegrate multiple segmentation algorithms in order to increase the robustness of the approach. In terms of the top-down feedback,\nthis is provided by two information sources: i) An estimation of the robot position that reduces the searching scope for potential\nmatches with previously selected landmarks, ii) A set of weights that, according to the results of previous recognitions, controls\nthe influence of each segmentation algorithm in the recognition of each landmark. We test our approach with encouraging results\nin three datasets corresponding to real-world scenarios.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/Pablo-RAS-08.pdf}\n}\n\n
\n
\n\n\n
\n In this paper, we tackle the problem of unsupervised selection and posterior recognition of visual landmarks in images sequences acquired by an indoor mobile robot. This is a highly valuable perceptual capability for a wide variety of robotic applications, in particular autonomous navigation. Our method combines a bottom-up data driven approach with top-down feedback provided by high level semantic representations. The bottom-up approach is based on three main mechanisms: visual attention, area segmentation, and landmark characterization. As there is no segmentation method that works properly in every situation, we integrate multiple segmentation algorithms in order to increase the robustness of the approach. In terms of the top-down feedback, this is provided by two information sources: i) An estimation of the robot position that reduces the searching scope for potential matches with previously selected landmarks, ii) A set of weights that, according to the results of previous recognitions, controls the influence of each segmentation algorithm in the recognition of each landmark. We test our approach with encouraging results in three datasets corresponding to real-world scenarios. \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Unsupervised Anomaly Detection in Large Databases Using Bayesian Networks.\n \n\n\n \n Cansado, A.; and Soto, A.\n\n\n \n\n\n\n Applied Artificial Intelligence, 22(4): 309-330. 2008.\n \n\n\n\n
\n\n\n\n \n \n \"UnsupervisedPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 2 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Cansado:Soto:2008,\n  Author = {A. Cansado and A. Soto},\n  Title = {Unsupervised Anomaly Detection in Large Databases Using Bayesian\nNetworks},\n  Journal = {Applied Artificial Intelligence},\n  Volume = {22},\n  Number = {4},\n  Pages = {309-330},\n  Year = {2008},\n    abstract = {Today, there has been a massive proliferation of huge databases storing\nvaluable information. The opportunities of an effective use of these new\ndata sources are enormous, however, the huge size and dimensionality of\ncurrent large databases call for new ideas to scale up current statistical\nand computational approaches. This paper presents an application of Ar-\ntificial Intelligence technology to the problem of automatic detection of\ncandidate anomalous records in a large database. We build our approach\nwith three main goals in mind: 1)An effective detection of the records\nthat are potentially anomalous, 2)A suitable selection of the subset of at-\ntributes that explains what makes a record anomalous, and 3)An efficient\nimplementation that allows us to scale the approach to large databases.\nOur algorithm, called Bayesian Network Anomaly Detector (BNAD), uses the joint probability density function (pdf) provided by a Bayesian Net-\nwork (BN) to achieve these goals. By using appropriate data structures,\nadvanced caching techniques, the flexibility of Gaussian Mixture models,\nand the efficiency of BNs to model joint pdfs, BNAD manages to effi-\nciently learn a suitable BN from a large dataset. We test BNAD using\nsynthetic and real databases, the latter from the fields of manufacturing\nand astronomy, obtaining encouraging results.\n},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/Cansado-Soto-AAI-2007.pdf}\n}\n\n
\n
\n\n\n
\n Today, there has been a massive proliferation of huge databases storing valuable information. The opportunities of an effective use of these new data sources are enormous, however, the huge size and dimensionality of current large databases call for new ideas to scale up current statistical and computational approaches. This paper presents an application of Ar- tificial Intelligence technology to the problem of automatic detection of candidate anomalous records in a large database. We build our approach with three main goals in mind: 1)An effective detection of the records that are potentially anomalous, 2)A suitable selection of the subset of at- tributes that explains what makes a record anomalous, and 3)An efficient implementation that allows us to scale the approach to large databases. Our algorithm, called Bayesian Network Anomaly Detector (BNAD), uses the joint probability density function (pdf) provided by a Bayesian Net- work (BN) to achieve these goals. By using appropriate data structures, advanced caching techniques, the flexibility of Gaussian Mixture models, and the efficiency of BNs to model joint pdfs, BNAD manages to effi- ciently learn a suitable BN from a large dataset. We test BNAD using synthetic and real databases, the latter from the fields of manufacturing and astronomy, obtaining encouraging results. \n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Real-Time Robot Localization In Indoor Environments Using Structural Information.\n \n\n\n \n Espinace, P.; Soto, A.; and Torres-Torriti, M.\n\n\n \n\n\n\n In IEEE Latin American Robotics Symposium (LARS), 2008. \n \n\n\n\n
\n\n\n\n \n \n \"Real-TimePaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Espinace:Soto:Torres:2008,\n  Author         = {P. Espinace and A. Soto and M. Torres-Torriti},\n  Title          = {Real-Time Robot Localization In Indoor Environments Using\nStructural Information},\n  booktitle   = {IEEE Latin American Robotics Symposium (LARS)},\n  pages = {},\n  year = {2008},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n
\n \n\n \n \n \n\n Improving the Selection and Detection of Visual Landmarks Through Object Tracking.\n \n\n\n \n Espinace, P.; and Soto, A.\n\n\n \n\n\n\n In IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), Workshop on Visual Localization for Mobile Platforms, 2008. \n \n\n\n\n
\n\n\n\n \n \n \"ImprovingPaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Espinace:Soto:2008b,\n  Author         = {P. Espinace and A. Soto},\n  Title          = {Improving the Selection and Detection of Visual Landmarks\nThrough Object Tracking},\n  booktitle   = {IEEE Conf. on Computer Vision and Pattern Recognition (CVPR),\nWorkshop on Visual Localization for Mobile Platforms},\n  pages = {},\n  year = {2008},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n
\n \n\n \n \n \n\n Human Detection in Indoor Environments Using Multiple Visual Cues and a Mobile Robot.\n \n\n\n \n Pszczolkowski, S.; and Soto, A.\n\n\n \n\n\n\n In Iberoamerican Congress on Pattern Recognition (CIARP), LNCS 4756, pages 350-359, 2008. \n \n\n\n\n
\n\n\n\n \n \n \"HumanPaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n  \n \n 2 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Pszczolkowski:Soto:2008,\n  Author         = {S. Pszczolkowski and A. Soto},\n  Title          = {Human Detection in Indoor Environments Using Multiple Visual\nCues and a Mobile Robot},\n  booktitle   = {Iberoamerican Congress on Pattern Recognition (CIARP), LNCS\n4756},\n  pages = {350-359},\n  year = {2008},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 2007\n \n \n (4)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n A Statistical approach to simultaneous mapping and localization for mobile robots.\n \n\n\n \n Araneda, A.; Fienberg, S.; and Soto, A.\n\n\n \n\n\n\n The Annals of Applied Statistics, 1(1): 66-84. 2007.\n \n\n\n\n
\n\n\n\n \n \n \"APaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Araneda:Fienberg:Soto:2007,\n  Author         = {A. Araneda and S. Fienberg and A. Soto},\n  Title          = {A Statistical approach to simultaneous mapping and\nlocalization for mobile robots},\n  journal      = {The Annals of Applied Statistics},\n  volume        = {1},\n  number        = {1},\n  pages         = {66-84},\n  year           = {2007},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n
\n \n\n \n \n \n\n Using Data Mining Techniques to Predict Industrial Wine Problem Fermentations.\n \n\n\n \n Urtubia; Perez-Correa, J. R.; Soto, A.; and Pszczolkowski, P.\n\n\n \n\n\n\n Food Control, 18(12): 1512-1517. 2007.\n \n\n\n\n
\n\n\n\n \n \n \"UsingPaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Urtubia:EtAl:2008,\n  Author         = {Urtubia and J. R. Perez-Correa and A. Soto and P.\nPszczolkowski},\n  Title          = {Using Data Mining Techniques to Predict Industrial Wine\nProblem Fermentations},\n  journal      = {Food Control},\n  volume        = {18},\n  number        = {12},\n  pages         = {1512-1517},\n  year           = {2007},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n
\n \n\n \n \n \n\n An Accelerated Algorithm for Density Estimation in Large Databases, Using Gaussian Mixtures.\n \n\n\n \n Soto, A.; Zavala, F.; and Araneda, A.\n\n\n \n\n\n\n Cybernetics and Systems: An International Journal, 38(2): 123-139. 2007.\n \n\n\n\n
\n\n\n\n \n \n \"AnPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 2 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Soto:Zavala:Araneda:2007,\n  Author         = {A. Soto and F. Zavala and A. Araneda},\n  Title          = {An Accelerated Algorithm for Density Estimation in Large\nDatabases, Using Gaussian Mixtures},\n  journal      = {Cybernetics and Systems: An International Journal},\n  volume        = {38},\n  number        = {2},\n  pages         = {123-139},\n  year           = {2007},\n  abstract = {Today, with the advances of computer storage and technology, there\nare huge datasets available, offering an opportunity to extract\nvaluable information. Probabilistic approaches are specially suited\nto learn from data by representing  knowledge as density functions.\nIn this paper, we choose  Gaussian Mixture Models (GMMs) to\nrepresent densities, as they possess great flexibility to adequate\n to a wide class of problems. The\nclassical estimation approach for GMMs corresponds to the iterative\nalgorithm of Expectation Maximization. This approach, however, does\nnot scale properly to meet the high demanding processing\nrequirements of large databases. In this paper we introduce an\nEM-based algorithm, that solves the scalability problem. Our\napproach is based on the concept of data condensation which, in\naddition to substantially diminishing the computational load,\nprovides sound starting values that allow the algorithm to reach\nconvergence faster. We also focus on the model selection problem. We\ntest our algorithm using synthetic and real databases, and find\nseveral advantages, when compared to other standard existing\nprocedures.},\n  url = {http://saturno.ing.puc.cl/media/papers_alvaro/Felipe-07.pdf}\n}\n\n
\n
\n\n\n
\n Today, with the advances of computer storage and technology, there are huge datasets available, offering an opportunity to extract valuable information. Probabilistic approaches are specially suited to learn from data by representing knowledge as density functions. In this paper, we choose Gaussian Mixture Models (GMMs) to represent densities, as they possess great flexibility to adequate to a wide class of problems. The classical estimation approach for GMMs corresponds to the iterative algorithm of Expectation Maximization. This approach, however, does not scale properly to meet the high demanding processing requirements of large databases. In this paper we introduce an EM-based algorithm, that solves the scalability problem. Our approach is based on the concept of data condensation which, in addition to substantially diminishing the computational load, provides sound starting values that allow the algorithm to reach convergence faster. We also focus on the model selection problem. We test our algorithm using synthetic and real databases, and find several advantages, when compared to other standard existing procedures.\n
\n\n\n
\n\n\n
\n \n\n \n \n \n\n Computer Vision for Quality Control in Latin American Food Industry, A Case Study.\n \n\n\n \n Aguilera, J.; Cipriano, A.; Erana, M.; Lillo, I.; Mery, D.; Soto, A.; and Valdivieso, C.\n\n\n \n\n\n\n In Int. Conf. on Computer Vision (ICCV): Workshop on Computer Vision Applications for Developing Countries, 2007. \n \n\n\n\n
\n\n\n\n \n \n \"ComputerPaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Aguilera:EtAt:2007,\n  Author         = {JM. Aguilera and A. Cipriano and M. Erana and I. Lillo and D. Mery and A.\nSoto and C. Valdivieso},\n  Title          = {Computer Vision for Quality Control in Latin American Food\nIndustry, A Case Study},\n  booktitle   = {Int. Conf. on Computer Vision (ICCV): Workshop on Computer\nVision Applications for Developing Countries},\n  pages = {},\n  year = {2007},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 2006\n \n \n (2)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n Automatic Selection and Detection of Visual Landmarks Using Multiple Segmentations.\n \n\n\n \n Langdon, D.; Soto, A.; and Mery, D.\n\n\n \n\n\n\n In IEEE Pacific-Rim Symposium on Image and Video Technology (PSIVT), LNCS 4319, pages 601-610, 2006. \n \n\n\n\n
\n\n\n\n \n \n \"AutomaticPaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Langdon:Soto:Mery:2006,\n  Author         = {D. Langdon and A. Soto and D. Mery},\n  Title          = {Automatic Selection and Detection of Visual Landmarks Using\nMultiple Segmentations},\n  booktitle   = {IEEE Pacific-Rim Symposium on Image and Video Technology\n(PSIVT), LNCS 4319},\n  pages = {601-610},\n  year = {2006},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n
\n \n\n \n \n \n\n A Mobile Robotics Course for Undergraduate Students in Computer Science.\n \n\n\n \n Soto, A.; Espinace, P.; and Mitnik, R.\n\n\n \n\n\n\n In IEEE Latin American Robotics Symposium, LARS, pages 187-192, 2006. \n \n\n\n\n
\n\n\n\n \n \n \"APaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Soto:Espinace:Mitnik:2006,\n  Author         = {A. Soto and P. Espinace and R. Mitnik},\n  Title          = {A Mobile Robotics Course for Undergraduate Students in\nComputer Science},\n  booktitle   = {IEEE Latin American Robotics Symposium, LARS},\n  pages = {187-192},\n  year = {2006},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 2005\n \n \n (2)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n Important Sampling in Mapping and Localization by a Mobile Robot.\n \n\n\n \n Araneda, A.; and Soto, A.\n\n\n \n\n\n\n In Workshop on Case Studies of Bayesian Statistics, 2005. \n \n\n\n\n
\n\n\n\n \n \n \"ImportantPaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Araneda:Soto:2005,\n  Author         = {A. Araneda and A. Soto},\n  Title          = {Important Sampling in Mapping and Localization by a Mobile\nRobot},\n  booktitle   = {Workshop on Case Studies of Bayesian Statistics},\n  year = {2005},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n
\n \n\n \n \n \n\n Self Adaptive Particle Filter.\n \n\n\n \n Soto, A.\n\n\n \n\n\n\n In Proceedings of International Join Conference on Artificial Intelligence (IJCAI), pages 1398-1406, 2005. \n \n\n\n\n
\n\n\n\n \n \n \"SelfPaper\n  \n \n\n \n\n bibtex\n \n\n \n  \n \n abstract \n \n\n \n  \n \n 3 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Soto:2005,\n  Author         = {A. Soto},\n  Title          = {Self Adaptive Particle Filter},\n  booktitle   = {Proceedings of International Join Conference on Artificial\nIntelligence (IJCAI)},\n  pages = {1398-1406},\n  year = {2005},\n    abstract = {The particle filter has emerged as a useful tool for\nproblems requiring dynamic state estimation. The\nefficiency and accuracy of the filter depend mostly\non the number of particles used in the estimation\nand on the propagation function used to re-allocate\nthese particles at each iteration. Both features are\nspecified beforehand and are kept fixed in the reg-\nular implementation of the filter. In practice this\nmay be highly inappropriate since it ignores errors\nin the models and the varying dynamics of the pro-\ncesses. This work presents a self adaptive version\nof the particle filter that uses statistical methods to\nadapt the number of particles and the propagation\nfunction at each iteration. Furthermore, our method\npresents similar computational load than the stan-\ndard particle filter. We show the advantages of the\nself adaptive filter by applying it to a synthetic ex-\nample and to the visual tracking of targets in a real\nvideo sequence.\n},\n  url = {Soto-IJCAI-0\t5.pdf}\n}\n\n
\n
\n\n\n
\n The particle filter has emerged as a useful tool for problems requiring dynamic state estimation. The efficiency and accuracy of the filter depend mostly on the number of particles used in the estimation and on the propagation function used to re-allocate these particles at each iteration. Both features are specified beforehand and are kept fixed in the reg- ular implementation of the filter. In practice this may be highly inappropriate since it ignores errors in the models and the varying dynamics of the pro- cesses. This work presents a self adaptive version of the particle filter that uses statistical methods to adapt the number of particles and the propagation function at each iteration. Furthermore, our method presents similar computational load than the stan- dard particle filter. We show the advantages of the self adaptive filter by applying it to a synthetic ex- ample and to the visual tracking of targets in a real video sequence. \n
\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 2004\n \n \n (4)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n Statistical Inference in Mapping and Localization for Mobile Robots.\n \n\n\n \n Araneda, A.; and A. Soto, undefined\n\n\n \n\n\n\n In Advances in Artificial Intelligence, Iberamia-04, LNAI 3315, pages 545-554, 2004. \n \n\n\n\n
\n\n\n\n \n \n \"StatisticalPaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Araneda:Soto:2004,\n  Author         = {A. Araneda and A. Soto,},\n  Title          = {Statistical Inference in Mapping and Localization for Mobile\nRobots},\n  booktitle   = {Advances in Artificial Intelligence, Iberamia-04, LNAI 3315},\n  pages = {545-554},\n  year = {2004},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n
\n \n\n \n \n \n\n Mobile Robotic Supported Collaborative Learning (MRSCL).\n \n\n\n \n Mitnik, R.; Nussbaum, M.; and Soto, A.\n\n\n \n\n\n\n In Advances in Artificial Intelligence, Iberamia-04, LNAI 3315, pages 912-921, 2004. \n \n\n\n\n
\n\n\n\n \n \n \"MobilePaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Mitnik:Nussbaum:Soto:2004,\n  Author         = {R. Mitnik and M. Nussbaum and A. Soto},\n  Title          = {Mobile Robotic Supported Collaborative Learning (MRSCL)},\n  booktitle   = {Advances in Artificial Intelligence, Iberamia-04, LNAI 3315},\n  pages = {912-921},\n  year = {2004},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n
\n \n\n \n \n \n\n Detection of Rare Objects in Massive Astrophysical Data Sets Using Innovative Knowledge Discovery Technology.\n \n\n\n \n Soto, A.; A.Cansado; and Zavala, F.\n\n\n \n\n\n\n In Astronomical Data Analysis Software & Systems Conf. Series (ADASS), pages 66-72, 2004. \n \n\n\n\n
\n\n\n\n \n \n \"DetectionPaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Soto:Cansado:Zavala:2008,\n  Author      = {A. Soto and A.Cansado and F. Zavala},\n  Title       = {Detection of Rare Objects in Massive Astrophysical Data Sets\nUsing Innovative Knowledge Discovery Technology},\n  booktitle   = {Astronomical Data Analysis Software & Systems Conf. Series\n(ADASS)},\n  pages = {66-72},\n  year = {2004},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n
\n \n\n \n \n \n\n A Method to Adaptively Propagate the Set of Samples Used by Particle Filters.\n \n\n\n \n Soto, A.\n\n\n \n\n\n\n In Lectures Notes in Artificial Intelligence, LNAI 3040, pages 47-56, 2004. \n \n\n\n\n
\n\n\n\n \n \n \"APaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Soto:2004,\n  Author         = {A. Soto},\n  Title          = {A Method to Adaptively Propagate the Set of Samples Used by\nParticle Filters},\n  booktitle   = {Lectures Notes in Artificial Intelligence, LNAI 3040},\n  pages = {47-56},\n  year = {2004},\n    abstract = {},\n  url = {}\n}\n\n\n
\n
\n\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 2003\n \n \n (4)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n Sequential Monte Carlo Methods for the Creation of Adaptive Software.\n \n\n\n \n Soto, A.; and Khosla, P.\n\n\n \n\n\n\n In 3th Int. Workshop on Self-adaptive Software, 2003. \n \n\n\n\n
\n\n\n\n \n \n \"SequentialPaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Soto:Khosla:2003a,\n  Author         = {A. Soto and P. Khosla},\n  Title          = {Sequential Monte Carlo Methods for the Creation of Adaptive\nSoftware},\n  booktitle   = {3th Int. Workshop on Self-adaptive Software},\n  pages = {},\n  year = {2003},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n
\n \n\n \n \n \n\n A Probabilistic Approach for Dynamic State Estimation Using Visual Information.\n \n\n\n \n Soto, A.; and Khosla, P.\n\n\n \n\n\n\n In Lectures Notes in Computer Science, LNCS 2821, pages 421-435, 2003. \n \n\n\n\n
\n\n\n\n \n \n \"APaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Soto:Khosla:2003b,\n  Author         = {A. Soto and P. Khosla},\n  Title          = {A Probabilistic Approach for Dynamic State Estimation Using\nVisual Information},\n  booktitle   = {Lectures Notes in Computer Science, LNCS 2821},\n  pages = {421-435},\n  year = {2003},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n
\n \n\n \n \n \n\n Adaptive Agent Based System for State Estimation Using Dynamic Multidimentional Information Sources.\n \n\n\n \n Soto, A.; and Khosla, P.\n\n\n \n\n\n\n In Lectures Notes in Computer Science, LNCS 2614, pages 66-83, 2003. \n \n\n\n\n
\n\n\n\n \n \n \"AdaptivePaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Soto:Khosla:2003c,\n  Author = {A. Soto and P. Khosla},\n  Title  = {Adaptive Agent Based System for State Estimation Using Dynamic\nMultidimentional Information Sources},\n  booktitle   = {Lectures Notes in Computer Science, LNCS 2614},\n  pages = {66-83},\n  year = {2003},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n
\n \n\n \n \n \n\n Probabilistic Adaptive Agent Based System for Dynamic State Estimation Using Multiple Visual Cues.\n \n\n\n \n Soto, A.; and Khosla, P.\n\n\n \n\n\n\n In volume 6, pages 559-572, 2003. \n \n\n\n\n
\n\n\n\n \n \n \"ProbabilisticPaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Soto:Khosla:2003d,\n  Author   = {A. Soto and P. Khosla},\n  Title    = {Probabilistic Adaptive Agent Based System for Dynamic State\nEstimation Using Multiple Visual Cues},\n  journal  = {Springer Tracts in Advanced Robotics (STAR)},\n  volume = {6},\n  pages = {559-572},\n  year = {2003},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 2002\n \n \n (2)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n A Probabilistic Approach for the Adaptive Integration of Multiple Visual Cues Using an Agent Framework.\n \n\n\n \n Soto, A.\n\n\n \n\n\n\n Technical Report PhD. Thesis, Robotics Institute, School of Computer Science, Tech Report CMU-RI-TR-02-30, Carnegie Mellon University, 2002.\n \n\n\n\n
\n\n\n\n \n \n \"APaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n  \n \n 7 downloads\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@techreport{Soto:2002,\n    author = {A. Soto},\n    title = {A Probabilistic Approach for the Adaptive Integration of Multiple\nVisual Cues Using an Agent Framework},\n    number = {PhD. Thesis, Robotics Institute, School of Computer Science, Tech\nReport CMU-RI-TR-02-30},\n    institution = {Carnegie Mellon University},\n    year = {2002},\n      abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n
\n \n\n \n \n \n\n Recent Advances in Distributed Tactical Surveillance.\n \n\n\n \n Saptharishi, M.; Bhat, K.; Diehl, C.; Oliver, S.; Savvides, M.; Soto, A.; Dolan, J.; and Khosla, P.\n\n\n \n\n\n\n In SPIE on Unattended Ground Sensor Technologies and Applications, Aerosense, 2002. \n \n\n\n\n
\n\n\n\n \n \n \"RecentPaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n  \n \n 1 download\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Saptharishi:EtAl:2002,\n  Author   = {M. Saptharishi and K. Bhat and C. Diehl and S. Oliver and M.\nSavvides and A. Soto and J. Dolan and P. Khosla},\n  Title    = {Recent Advances in Distributed Tactical Surveillance},\n  booktitle   = {SPIE on Unattended Ground Sensor Technologies and Applications,\nAerosense},\n  pages = {},\n  year = {2002},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 1999\n \n \n (3)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n CyberATVs: Dynamic and Distributed Reconnaissance and Surveillance Using All Terrain UGVs.\n \n\n\n \n Soto, A.; Saptharishi, M.; Dolan, J.; Trebi-Ollennu, A.; and Khosla, P.\n\n\n \n\n\n\n In Proceedings of the International Conference on Field and Service Robotics (FSR), 1999. \n \n\n\n\n
\n\n\n\n \n \n \"CyberATVs:Paper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Soto:EtAl:2002,\n  Author   = {A. Soto and M. Saptharishi and J. Dolan and A. Trebi-Ollennu and\nP. Khosla},\n  Title    = {CyberATVs: Dynamic and Distributed Reconnaissance and Surveillance\nUsing All Terrain UGVs},\n  booktitle   = {Proceedings of the International Conference on Field and\nService Robotics (FSR)},\n  pages = {},\n  year = {1999},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n
\n \n\n \n \n \n\n An Effective Mobile Robot Educator with a Full-Time Job.\n \n\n\n \n Nourbakhsh, I.; Bobenage, J.; Grange, S.; Lutz, R.; Meyer, R.; and Soto, A.\n\n\n \n\n\n\n Artificial Intelligence, 114(1-2): 95-124. 1999.\n \n\n\n\n
\n\n\n\n \n \n \"AnPaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Nourbakhsh:EtAl:1999,\n  Author         = {I. Nourbakhsh and J. Bobenage and S. Grange and R. Lutz and\nR. Meyer and A. Soto},\n  Title          = {An Effective Mobile Robot Educator with a Full-Time Job},\n  journal      = {Artificial Intelligence},\n  volume        = {114},\n  number        = {1-2},\n  pages         = {95-124},\n  year           = {1999},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n
\n \n\n \n \n \n\n Distributed Tactical Surveillance with ATVs.\n \n\n\n \n Dolan, J.; Trebi-Ollennu, A.; Soto, A.; and Khosla, P.\n\n\n \n\n\n\n In SPIE on Unattended Ground Sensor Technologies and Applications, Aerosense, Vol. 3693, 1999. \n \n\n\n\n
\n\n\n\n \n \n \"DistributedPaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Dolan:EtAl:1999,\n  Author   = {J. Dolan and A. Trebi-Ollennu and A. Soto and P. Khosla},\n  Title    = {Distributed Tactical Surveillance with ATVs},\n  booktitle   = {SPIE on Unattended Ground Sensor Technologies and Applications,\nAerosense, Vol. 3693},\n  pages = {},\n  year = {1999},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 1998\n \n \n (2)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n A Scenario for Planning Visual Navigation of a Mobile Robot.\n \n\n\n \n Soto, A.; and Nourbakhsh, I.\n\n\n \n\n\n\n In American Association for Artificial Intelligence (AAAI), Fall Symposium Series, 1998. \n \n\n\n\n
\n\n\n\n \n \n \"APaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Soto:Illah:1998,\n  Author   = {A. Soto and I. Nourbakhsh},\n  Title    = {A Scenario for Planning Visual Navigation of a Mobile Robot},\n  booktitle   = {American Association for Artificial Intelligence (AAAI), Fall\nSymposium Series},\n  pages = {},\n  year = {1998},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n
\n \n\n \n \n \n\n A Real Time Visual Sensor for Supervision of Flotations Cells.\n \n\n\n \n Cipriano, A.; Guarini, M.; Vidal, R.; Soto, A.; Sepúlveda, C.; Mery, D.; and Briseño, H.\n\n\n \n\n\n\n Minerals Engineering, 11(6): 489-499. 1998.\n \n\n\n\n
\n\n\n\n \n \n \"APaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@article{Cipriano:EtAl:1998,\n  Author   = {A. Cipriano and M. Guarini and R. Vidal and A. Soto and C.\nSepúlveda and D. Mery and H. Briseño},\n  Title    = {A Real Time Visual Sensor for Supervision of Flotations Cells},\n  journal   = {Minerals Engineering},\n  volume        = {11},\n  number        = {6},\n  pages = {489-499},\n  year = {1998},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 1997\n \n \n (1)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n Expert supervision of flotation cells using digital image processing.\n \n\n\n \n Cipriano, A.; Guarini, M.; Soto, A.; Briseño, H.; and D. Mery, undefined\n\n\n \n\n\n\n In In Proc. of 20th Int. Mineral Processing Congress, pages 281-292, 1997. \n \n\n\n\n
\n\n\n\n \n \n \"ExpertPaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Cipriano:Et:Al:1997,\n  Author   = {A. Cipriano and M. Guarini and A. Soto and H. Briseño and D.\nMery,},\n  Title    = {Expert supervision of flotation cells using digital image\nprocessing},\n  booktitle   = {In Proc. of 20th Int. Mineral Processing Congress},\n  pages = {281-292},\n  year = {1997},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 1996\n \n \n (1)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n Image processing applied to real time measurement of traffic flow.\n \n\n\n \n Soto, A.; and Cipriano, A.\n\n\n \n\n\n\n In In Proc. of 28th Southeastern Symposium on System Theory, 1996. \n \n\n\n\n
\n\n\n\n \n \n \"ImagePaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Soto:Cipriano:1996,\n  Author   = {A. Soto and A. Cipriano},\n  Title    = {Image processing applied to real time measurement of traffic\nflow},\n  booktitle   = {In Proc. of 28th Southeastern Symposium on System Theory},\n  pages = {},\n  year = {1996},\n    abstract = {},\n  url = {}\n}\n\n
\n
\n\n\n\n
\n\n\n\n\n\n
\n
\n\n
\n
\n  \n 1995\n \n \n (1)\n \n \n
\n
\n \n \n
\n \n\n \n \n \n\n Measurement of physical characteristics of foam in flotation cells.\n \n\n\n \n Guarini, M.; Soto, A.; Cipriano, A.; Guesalaga, A.; and Caceres, J.\n\n\n \n\n\n\n In In Proc. of Int. Conference: Copper-95, 1995. \n \n\n\n\n
\n\n\n\n \n \n \"MeasurementPaper\n  \n \n\n \n\n bibtex\n \n\n \n\n \n  \n \n 1 download\n \n \n\n \n \n \n \n \n \n \n\n\n\n
\n
@inproceedings{Guarini:EtAl:1995,\n  Author   = {M. Guarini and A. Soto and A. Cipriano and A. Guesalaga and J. Caceres},\n  Title    = {Measurement of physical characteristics of foam in flotation cells},\n  booktitle   = {In Proc. of Int. Conference: Copper-95},\n  pages = {},\n  year = {1995},\n    abstract = {},\n  url = {}\n}\n
\n
\n\n\n\n
\n\n\n\n\n\n
\n
\n\n\n\n\n
\n\n\n \n
\n
\n \n

Embedding in another Page

\n
\n
\n\n

\n Copy&paste any of the following snippets into an existing\n page to embed this page. For more details see the\n documention.\n

\n\n JavaScript\n (easiest)\n
\n \n <script src=\"https://bibbase.org/show?bib=http://saturno.ing.puc.cl/media/bib_files/actualizado_cgtercQ.bib&jsonp=1&jsonp=1\"></script>\n \n
\n\n PHP\n
\n \n <?php\n $contents = file_get_contents(\"https://bibbase.org/show?bib=http://saturno.ing.puc.cl/media/bib_files/actualizado_cgtercQ.bib&jsonp=1\");\n print_r($contents);\n ?>\n \n
\n\n iFrame\n (not recommended)\n
\n \n <iframe src=\"https://bibbase.org/show?bib=http://saturno.ing.puc.cl/media/bib_files/actualizado_cgtercQ.bib&jsonp=1\"></iframe>\n \n
\n\n\n
\n
\n \n
\n
\n\n \n \n \n \n\n
\n"}; document.write(bibbase_data.data);