Lossy intermediate deep learning feature compression and evaluation
With the unprecedented success of deep learning in computer vision tasks, many cloud-based visual analysis applications are powered by deep learning models. However, the deep learning models are also characterized with high computational complexity and are task-specific, which may hinder the large-s...
Saved in:
Main Authors: | , , , , , |
---|---|
Other Authors: | |
Format: | Conference or Workshop Item |
Language: | English |
Published: |
2020
|
Subjects: | |
Online Access: | https://hdl.handle.net/10356/144189 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
id |
sg-ntu-dr.10356-144189 |
---|---|
record_format |
dspace |
spelling |
sg-ntu-dr.10356-1441892020-10-20T02:27:50Z Lossy intermediate deep learning feature compression and evaluation Chen, Zhuo Fan, Kui Wang, Shiqi Duan, Ling-Yu Lin, Weisi Kot, Alex Interdisciplinary Graduate School (IGS) 27th ACM International Conference on Multimedia Engineering::Computer science and engineering Feature Compression Deep Learning With the unprecedented success of deep learning in computer vision tasks, many cloud-based visual analysis applications are powered by deep learning models. However, the deep learning models are also characterized with high computational complexity and are task-specific, which may hinder the large-scale implementation of the conventional data communication paradigms. To enable a better balance among bandwidth usage, computational load and the generalization capability for cloud-end servers, we propose to compress and transmit intermediate deep learning features instead of visual signals and ultimately utilized features. The proposed strategy also provides a promising way for the standardization of deep feature coding. As the first attempt to this problem, we present a lossy compression framework and evaluation metrics for intermediate deep feature compression. Comprehensive experimental results show the effectiveness of our proposed methods and the feasibility of the proposed data transmission strategy. It is worth mentioning that the proposed compression framework and evaluation metrics have been adopted into the ongoing AVS (Audio Video Coding Standard Workgroup) - Visual Feature Coding Standard. Ministry of Education (MOE) National Research Foundation (NRF) Accepted version This research is supported by the NTU-PKU Joint Research Institute, a collaboration between the Nanyang Technological University (NTU), Singapore, and Peking University (PKU), China, which is sponsored by a donation from the Ng Teng Fong Charitable Foundation. The research work was done at the Rapid-Rich Object Search (ROSE) Lab at NTU. This work is also supported in part by Singapore Ministry of Education Tier-2 Fund MOE2016-T2-2-057(S), in part by the National Natural Science Foundation of China under Grant 61661146005 and Grant U1611461, in part by Hong Kong RGC Early Career Scheme 9048122 (CityU 21211018), in part by City University of Hong Kong under Grant 7200539/CS, and in part by the National Research Foundation, Prime Minister’s Office, Singapore, through the NRF-NSFC Grant, under Grant NRF2016NRF-NSFC001-098. 2020-10-20T02:25:26Z 2020-10-20T02:25:26Z 2019 Conference Paper Chen, Z., Fan, K., Wang, S., Duan, L.-Y., Lin, W., & Kot, A. (2019). Lossy intermediate deep learning feature compression and evaluation. Proceedings of 27th ACM International Conference on Multimedia, 2414- 2422. doi:10.1145/3343031.3350849 9781450368896 https://hdl.handle.net/10356/144189 10.1145/3343031.3350849 2414 2422 en © 2019 Association for Computing Machinery (ACM). All rights reserved. This paper was published in 27th ACM International Conference on Multimedia and is made available with permission of Association for Computing Machinery (ACM). application/pdf |
institution |
Nanyang Technological University |
building |
NTU Library |
continent |
Asia |
country |
Singapore Singapore |
content_provider |
NTU Library |
collection |
DR-NTU |
language |
English |
topic |
Engineering::Computer science and engineering Feature Compression Deep Learning |
spellingShingle |
Engineering::Computer science and engineering Feature Compression Deep Learning Chen, Zhuo Fan, Kui Wang, Shiqi Duan, Ling-Yu Lin, Weisi Kot, Alex Lossy intermediate deep learning feature compression and evaluation |
description |
With the unprecedented success of deep learning in computer vision tasks, many cloud-based visual analysis applications are powered by deep learning models. However, the deep learning models are also characterized with high computational complexity and are task-specific, which may hinder the large-scale implementation of the conventional data communication paradigms. To enable a better balance among bandwidth usage, computational load and the generalization capability for cloud-end servers, we propose to compress and transmit intermediate deep learning features instead of visual signals and ultimately utilized features. The proposed strategy also provides a promising way for the standardization of deep feature coding. As the first attempt to this problem, we present a lossy compression framework and evaluation metrics for intermediate deep feature compression. Comprehensive experimental results show the effectiveness of our proposed methods and the feasibility of the proposed data transmission strategy. It is worth mentioning that the proposed compression framework and evaluation metrics have been adopted into the ongoing AVS (Audio Video Coding Standard Workgroup) - Visual Feature Coding Standard. |
author2 |
Interdisciplinary Graduate School (IGS) |
author_facet |
Interdisciplinary Graduate School (IGS) Chen, Zhuo Fan, Kui Wang, Shiqi Duan, Ling-Yu Lin, Weisi Kot, Alex |
format |
Conference or Workshop Item |
author |
Chen, Zhuo Fan, Kui Wang, Shiqi Duan, Ling-Yu Lin, Weisi Kot, Alex |
author_sort |
Chen, Zhuo |
title |
Lossy intermediate deep learning feature compression and evaluation |
title_short |
Lossy intermediate deep learning feature compression and evaluation |
title_full |
Lossy intermediate deep learning feature compression and evaluation |
title_fullStr |
Lossy intermediate deep learning feature compression and evaluation |
title_full_unstemmed |
Lossy intermediate deep learning feature compression and evaluation |
title_sort |
lossy intermediate deep learning feature compression and evaluation |
publishDate |
2020 |
url |
https://hdl.handle.net/10356/144189 |
_version_ |
1683492951192240128 |