A Deep-Unfolded Reference-Based RPCA Network For Video Foreground-Background Separation. Van Luong, H., Joukovsky, B., Eldar, Y. C., & Deligiannis, N. In 2020 28th European Signal Processing Conference (EUSIPCO), pages 1432-1436, Aug, 2020. Paper doi abstract bibtex Deep unfolded neural networks are designed by unrolling the iterations of optimization algorithms. They can be shown to achieve faster convergence and higher accuracy than their optimization counterparts. This paper proposes a new deep-unfolding-based network design for the problem of Robust Principal Component Analysis (RPCA) with application to video foreground-background separation. Unlike existing designs, our approach focuses on modeling the temporal correlation between the sparse representations of consecutive video frames. To this end, we perform the unfolding of an iterative algorithm for solving reweighted ℓ1-ℓ1 minimization; this unfolding leads to a different proximal operator (a.k.a. different activation function) adaptively learned per neuron. Experimentation using the moving MNIST dataset shows that the proposed network outperforms a recently proposed state-of-the-art RPCA network in the task of video foreground-background separation.
@InProceedings{9287416,
author = {H. {Van Luong} and B. Joukovsky and Y. C. Eldar and N. Deligiannis},
booktitle = {2020 28th European Signal Processing Conference (EUSIPCO)},
title = {A Deep-Unfolded Reference-Based RPCA Network For Video Foreground-Background Separation},
year = {2020},
pages = {1432-1436},
abstract = {Deep unfolded neural networks are designed by unrolling the iterations of optimization algorithms. They can be shown to achieve faster convergence and higher accuracy than their optimization counterparts. This paper proposes a new deep-unfolding-based network design for the problem of Robust Principal Component Analysis (RPCA) with application to video foreground-background separation. Unlike existing designs, our approach focuses on modeling the temporal correlation between the sparse representations of consecutive video frames. To this end, we perform the unfolding of an iterative algorithm for solving reweighted ℓ1-ℓ1 minimization; this unfolding leads to a different proximal operator (a.k.a. different activation function) adaptively learned per neuron. Experimentation using the moving MNIST dataset shows that the proposed network outperforms a recently proposed state-of-the-art RPCA network in the task of video foreground-background separation.},
keywords = {Correlation;Neurons;Signal processing algorithms;Signal processing;Minimization;Task analysis;Optimization;Deep unfolding;deep learning;robust PCA;video analysis;foreground-background separation},
doi = {10.23919/Eusipco47968.2020.9287416},
issn = {2076-1465},
month = {Aug},
url = {https://www.eurasip.org/proceedings/eusipco/eusipco2020/pdfs/0001432.pdf},
}
Downloads: 0
{"_id":"EfFpftuXMRDZDuoFc","bibbaseid":"vanluong-joukovsky-eldar-deligiannis-adeepunfoldedreferencebasedrpcanetworkforvideoforegroundbackgroundseparation-2020","authorIDs":[],"author_short":["Van Luong, H.","Joukovsky, B.","Eldar, Y. C.","Deligiannis, N."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","author":[{"firstnames":["H."],"propositions":[],"lastnames":["Van Luong"],"suffixes":[]},{"firstnames":["B."],"propositions":[],"lastnames":["Joukovsky"],"suffixes":[]},{"firstnames":["Y.","C."],"propositions":[],"lastnames":["Eldar"],"suffixes":[]},{"firstnames":["N."],"propositions":[],"lastnames":["Deligiannis"],"suffixes":[]}],"booktitle":"2020 28th European Signal Processing Conference (EUSIPCO)","title":"A Deep-Unfolded Reference-Based RPCA Network For Video Foreground-Background Separation","year":"2020","pages":"1432-1436","abstract":"Deep unfolded neural networks are designed by unrolling the iterations of optimization algorithms. They can be shown to achieve faster convergence and higher accuracy than their optimization counterparts. This paper proposes a new deep-unfolding-based network design for the problem of Robust Principal Component Analysis (RPCA) with application to video foreground-background separation. Unlike existing designs, our approach focuses on modeling the temporal correlation between the sparse representations of consecutive video frames. To this end, we perform the unfolding of an iterative algorithm for solving reweighted ℓ1-ℓ1 minimization; this unfolding leads to a different proximal operator (a.k.a. different activation function) adaptively learned per neuron. Experimentation using the moving MNIST dataset shows that the proposed network outperforms a recently proposed state-of-the-art RPCA network in the task of video foreground-background separation.","keywords":"Correlation;Neurons;Signal processing algorithms;Signal processing;Minimization;Task analysis;Optimization;Deep unfolding;deep learning;robust PCA;video analysis;foreground-background separation","doi":"10.23919/Eusipco47968.2020.9287416","issn":"2076-1465","month":"Aug","url":"https://www.eurasip.org/proceedings/eusipco/eusipco2020/pdfs/0001432.pdf","bibtex":"@InProceedings{9287416,\n author = {H. {Van Luong} and B. Joukovsky and Y. C. Eldar and N. Deligiannis},\n booktitle = {2020 28th European Signal Processing Conference (EUSIPCO)},\n title = {A Deep-Unfolded Reference-Based RPCA Network For Video Foreground-Background Separation},\n year = {2020},\n pages = {1432-1436},\n abstract = {Deep unfolded neural networks are designed by unrolling the iterations of optimization algorithms. They can be shown to achieve faster convergence and higher accuracy than their optimization counterparts. This paper proposes a new deep-unfolding-based network design for the problem of Robust Principal Component Analysis (RPCA) with application to video foreground-background separation. Unlike existing designs, our approach focuses on modeling the temporal correlation between the sparse representations of consecutive video frames. To this end, we perform the unfolding of an iterative algorithm for solving reweighted ℓ1-ℓ1 minimization; this unfolding leads to a different proximal operator (a.k.a. different activation function) adaptively learned per neuron. Experimentation using the moving MNIST dataset shows that the proposed network outperforms a recently proposed state-of-the-art RPCA network in the task of video foreground-background separation.},\n keywords = {Correlation;Neurons;Signal processing algorithms;Signal processing;Minimization;Task analysis;Optimization;Deep unfolding;deep learning;robust PCA;video analysis;foreground-background separation},\n doi = {10.23919/Eusipco47968.2020.9287416},\n issn = {2076-1465},\n month = {Aug},\n url = {https://www.eurasip.org/proceedings/eusipco/eusipco2020/pdfs/0001432.pdf},\n}\n\n","author_short":["Van Luong, H.","Joukovsky, B.","Eldar, Y. C.","Deligiannis, N."],"key":"9287416","id":"9287416","bibbaseid":"vanluong-joukovsky-eldar-deligiannis-adeepunfoldedreferencebasedrpcanetworkforvideoforegroundbackgroundseparation-2020","role":"author","urls":{"Paper":"https://www.eurasip.org/proceedings/eusipco/eusipco2020/pdfs/0001432.pdf"},"keyword":["Correlation;Neurons;Signal processing algorithms;Signal processing;Minimization;Task analysis;Optimization;Deep unfolding;deep learning;robust PCA;video analysis;foreground-background separation"],"metadata":{"authorlinks":{}},"downloads":0},"bibtype":"inproceedings","biburl":"https://raw.githubusercontent.com/Roznn/EUSIPCO/main/eusipco2020url.bib","creationDate":"2021-02-13T19:41:51.304Z","downloads":0,"keywords":["correlation;neurons;signal processing algorithms;signal processing;minimization;task analysis;optimization;deep unfolding;deep learning;robust pca;video analysis;foreground-background separation"],"search_terms":["deep","unfolded","reference","based","rpca","network","video","foreground","background","separation","van luong","joukovsky","eldar","deligiannis"],"title":"A Deep-Unfolded Reference-Based RPCA Network For Video Foreground-Background Separation","year":2020,"dataSources":["wXzutN6o5hxayPKdC","NBHz6C7PWuqwYyaqa"]}