Impact of temporal subsampling on accuracy and performance in practical video classification. Scheidegger, F., Cavigelli, L., Schaffner, M., Malossi, A. C. I., Bekas, C., & Benini, L. In 2017 25th European Signal Processing Conference (EUSIPCO), pages 996-1000, Aug, 2017.
![Impact of temporal subsampling on accuracy and performance in practical video classification [pdf]](https://bibbase.org/img/filetypes/pdf.svg "pdf")
Paper doi abstract bibtex
Paper doi abstract bibtex In this paper we evaluate three state-of-the-art neural-network-based approaches for large-scale video classification, where the computational efficiency of the inference step is of particular importance due to the ever increasing amount of data throughput for video streams. Our evaluation focuses on finding good efficiency vs. accuracy tradeoffs by evaluating different network configurations and parameterizations. In particular, we investigate the use of different temporal subsampling strategies, and show that they can be used to effectively trade computational workload against classification accuracy. Using a subset of the YouTube-8M dataset, we demonstrate that workload reductions in the order of 10×, 50× and 100× can be achieved with accuracy reductions of only 1.3%, 6.2% and 10.8%, respectively. Our results show that temporal subsampling is a simple and generic approach that behaves consistently over the considered classification pipelines and which does not require retraining of the underlying networks.
@InProceedings{8081357,
author = {F. Scheidegger and L. Cavigelli and M. Schaffner and A. C. I. Malossi and C. Bekas and L. Benini},
booktitle = {2017 25th European Signal Processing Conference (EUSIPCO)},
title = {Impact of temporal subsampling on accuracy and performance in practical video classification},
year = {2017},
pages = {996-1000},
abstract = {In this paper we evaluate three state-of-the-art neural-network-based approaches for large-scale video classification, where the computational efficiency of the inference step is of particular importance due to the ever increasing amount of data throughput for video streams. Our evaluation focuses on finding good efficiency vs. accuracy tradeoffs by evaluating different network configurations and parameterizations. In particular, we investigate the use of different temporal subsampling strategies, and show that they can be used to effectively trade computational workload against classification accuracy. Using a subset of the YouTube-8M dataset, we demonstrate that workload reductions in the order of 10×, 50× and 100× can be achieved with accuracy reductions of only 1.3%, 6.2% and 10.8%, respectively. Our results show that temporal subsampling is a simple and generic approach that behaves consistently over the considered classification pipelines and which does not require retraining of the underlying networks.},
keywords = {image classification;learning (artificial intelligence);neural nets;video signal processing;video streaming;YouTube-8M dataset;neural-network;large-scale video classification;computational efficiency;inference step;video streams;temporal subsampling strategies;classification pipelines;Training;Feature extraction;Video sequences;Artificial neural networks;Europe;Graphics processing units},
doi = {10.23919/EUSIPCO.2017.8081357},
issn = {2076-1465},
month = {Aug},
url = {https://www.eurasip.org/proceedings/eusipco/eusipco2017/papers/1570342773.pdf},
}Downloads: 0
{"_id":"fiuP2qAw3bLiNpP9s","bibbaseid":"scheidegger-cavigelli-schaffner-malossi-bekas-benini-impactoftemporalsubsamplingonaccuracyandperformanceinpracticalvideoclassification-2017","authorIDs":[],"author_short":["Scheidegger, F.","Cavigelli, L.","Schaffner, M.","Malossi, A. C. I.","Bekas, C.","Benini, L."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","author":[{"firstnames":["F."],"propositions":[],"lastnames":["Scheidegger"],"suffixes":[]},{"firstnames":["L."],"propositions":[],"lastnames":["Cavigelli"],"suffixes":[]},{"firstnames":["M."],"propositions":[],"lastnames":["Schaffner"],"suffixes":[]},{"firstnames":["A.","C.","I."],"propositions":[],"lastnames":["Malossi"],"suffixes":[]},{"firstnames":["C."],"propositions":[],"lastnames":["Bekas"],"suffixes":[]},{"firstnames":["L."],"propositions":[],"lastnames":["Benini"],"suffixes":[]}],"booktitle":"2017 25th European Signal Processing Conference (EUSIPCO)","title":"Impact of temporal subsampling on accuracy and performance in practical video classification","year":"2017","pages":"996-1000","abstract":"In this paper we evaluate three state-of-the-art neural-network-based approaches for large-scale video classification, where the computational efficiency of the inference step is of particular importance due to the ever increasing amount of data throughput for video streams. Our evaluation focuses on finding good efficiency vs. accuracy tradeoffs by evaluating different network configurations and parameterizations. In particular, we investigate the use of different temporal subsampling strategies, and show that they can be used to effectively trade computational workload against classification accuracy. Using a subset of the YouTube-8M dataset, we demonstrate that workload reductions in the order of 10×, 50× and 100× can be achieved with accuracy reductions of only 1.3%, 6.2% and 10.8%, respectively. Our results show that temporal subsampling is a simple and generic approach that behaves consistently over the considered classification pipelines and which does not require retraining of the underlying networks.","keywords":"image classification;learning (artificial intelligence);neural nets;video signal processing;video streaming;YouTube-8M dataset;neural-network;large-scale video classification;computational efficiency;inference step;video streams;temporal subsampling strategies;classification pipelines;Training;Feature extraction;Video sequences;Artificial neural networks;Europe;Graphics processing units","doi":"10.23919/EUSIPCO.2017.8081357","issn":"2076-1465","month":"Aug","url":"https://www.eurasip.org/proceedings/eusipco/eusipco2017/papers/1570342773.pdf","bibtex":"@InProceedings{8081357,\n author = {F. Scheidegger and L. Cavigelli and M. Schaffner and A. C. I. Malossi and C. Bekas and L. Benini},\n booktitle = {2017 25th European Signal Processing Conference (EUSIPCO)},\n title = {Impact of temporal subsampling on accuracy and performance in practical video classification},\n year = {2017},\n pages = {996-1000},\n abstract = {In this paper we evaluate three state-of-the-art neural-network-based approaches for large-scale video classification, where the computational efficiency of the inference step is of particular importance due to the ever increasing amount of data throughput for video streams. Our evaluation focuses on finding good efficiency vs. accuracy tradeoffs by evaluating different network configurations and parameterizations. In particular, we investigate the use of different temporal subsampling strategies, and show that they can be used to effectively trade computational workload against classification accuracy. Using a subset of the YouTube-8M dataset, we demonstrate that workload reductions in the order of 10×, 50× and 100× can be achieved with accuracy reductions of only 1.3%, 6.2% and 10.8%, respectively. Our results show that temporal subsampling is a simple and generic approach that behaves consistently over the considered classification pipelines and which does not require retraining of the underlying networks.},\n keywords = {image classification;learning (artificial intelligence);neural nets;video signal processing;video streaming;YouTube-8M dataset;neural-network;large-scale video classification;computational efficiency;inference step;video streams;temporal subsampling strategies;classification pipelines;Training;Feature extraction;Video sequences;Artificial neural networks;Europe;Graphics processing units},\n doi = {10.23919/EUSIPCO.2017.8081357},\n issn = {2076-1465},\n month = {Aug},\n url = {https://www.eurasip.org/proceedings/eusipco/eusipco2017/papers/1570342773.pdf},\n}\n\n","author_short":["Scheidegger, F.","Cavigelli, L.","Schaffner, M.","Malossi, A. C. I.","Bekas, C.","Benini, L."],"key":"8081357","id":"8081357","bibbaseid":"scheidegger-cavigelli-schaffner-malossi-bekas-benini-impactoftemporalsubsamplingonaccuracyandperformanceinpracticalvideoclassification-2017","role":"author","urls":{"Paper":"https://www.eurasip.org/proceedings/eusipco/eusipco2017/papers/1570342773.pdf"},"keyword":["image classification;learning (artificial intelligence);neural nets;video signal processing;video streaming;YouTube-8M dataset;neural-network;large-scale video classification;computational efficiency;inference step;video streams;temporal subsampling strategies;classification pipelines;Training;Feature extraction;Video sequences;Artificial neural networks;Europe;Graphics processing units"],"metadata":{"authorlinks":{}},"downloads":0},"bibtype":"inproceedings","biburl":"https://raw.githubusercontent.com/Roznn/EUSIPCO/main/eusipco2017url.bib","creationDate":"2021-02-13T16:38:25.621Z","downloads":0,"keywords":["image classification;learning (artificial intelligence);neural nets;video signal processing;video streaming;youtube-8m dataset;neural-network;large-scale video classification;computational efficiency;inference step;video streams;temporal subsampling strategies;classification pipelines;training;feature extraction;video sequences;artificial neural networks;europe;graphics processing units"],"search_terms":["impact","temporal","subsampling","accuracy","performance","practical","video","classification","scheidegger","cavigelli","schaffner","malossi","bekas","benini"],"title":"Impact of temporal subsampling on accuracy and performance in practical video classification","year":2017,"dataSources":["2MNbFYjMYTD6z7ExY","uP2aT6Qs8sfZJ6s8b"]}