PASS Patch Automatic Skip Scheme for Efficient On-Device Video Perception

PASS : Patch Automatic Skip Scheme for Efficient On-Device Video Perception

Real-time video perception tasks are often challenging on resource-constrained edge devices due to the issues of accuracy drop and hardware overhead, where saving computations is the key to performance improvement. Existing methods either rely on domain-specific neural chips or priorly searched mode...

Ausführliche Beschreibung

Bibliographische Detailangaben
Veröffentlicht in:IEEE transactions on pattern analysis and machine intelligence. - 1979. - 46(2024), 5 vom: 03. Apr., Seite 3938-3954
1. Verfasser: Zhou, Qihua (VerfasserIn)
Weitere Verfasser: Guo, Song, Pan, Jun, Liang, Jiacheng, Guo, Jingcai, Xu, Zhenda, Zhou, Jingren
Format: Online-Aufsatz
Sprache:English
Veröffentlicht: 2024
Zugriff auf das übergeordnete Werk:IEEE transactions on pattern analysis and machine intelligence
Schlagworte:Journal Article
LEADER 01000caa a22002652 4500
001 NLM366812882
003 DE-627
005 20240404234550.0
007 cr uuu---uuuuu
008 240114s2024 xx |||||o 00| ||eng c
024 7 |a 10.1109/TPAMI.2024.3350380  |2 doi 
028 5 2 |a pubmed24n1364.xml 
035 |a (DE-627)NLM366812882 
035 |a (NLM)38190691 
040 |a DE-627  |b ger  |c DE-627  |e rakwb 
041 |a eng 
100 1 |a Zhou, Qihua  |e verfasserin  |4 aut 
245 1 0 |a PASS  |b Patch Automatic Skip Scheme for Efficient On-Device Video Perception 
264 1 |c 2024 
336 |a Text  |b txt  |2 rdacontent 
337 |a ƒaComputermedien  |b c  |2 rdamedia 
338 |a ƒa Online-Ressource  |b cr  |2 rdacarrier 
500 |a Date Revised 03.04.2024 
500 |a published: Print-Electronic 
500 |a Citation Status PubMed-not-MEDLINE 
520 |a Real-time video perception tasks are often challenging on resource-constrained edge devices due to the issues of accuracy drop and hardware overhead, where saving computations is the key to performance improvement. Existing methods either rely on domain-specific neural chips or priorly searched models, which require specialized optimization according to different task properties. These limitations motivate us to design a general and task-independent methodology, called Patch Automatic Skip Scheme (PASS), which supports diverse video perception settings by decoupling acceleration and tasks. The gist is to capture inter-frame correlations and skip redundant computations at patch level, where the patch is a non-overlapping square block in visual. PASS equips each convolution layer with a learnable gate to selectively determine which patches could be safely skipped without degrading model accuracy. Specifically, we are the first to construct a self-supervisory procedure for gate optimization, which learns to extract contrastive representations from frame sequences. The pre-trained gates can serve as plug-and-play modules to implement patch-skippable neural backbones, and automatically generate proper skip strategy to accelerate different video-based downstream tasks, e.g., outperforming state-of-the-art MobileHumanPose in 3D pose estimation and FairMOT in multiple object tracking, by up to 9.43 × and 12.19 × speedups, respectively, on NVIDIA Jetson Nano devices 
650 4 |a Journal Article 
700 1 |a Guo, Song  |e verfasserin  |4 aut 
700 1 |a Pan, Jun  |e verfasserin  |4 aut 
700 1 |a Liang, Jiacheng  |e verfasserin  |4 aut 
700 1 |a Guo, Jingcai  |e verfasserin  |4 aut 
700 1 |a Xu, Zhenda  |e verfasserin  |4 aut 
700 1 |a Zhou, Jingren  |e verfasserin  |4 aut 
773 0 8 |i Enthalten in  |t IEEE transactions on pattern analysis and machine intelligence  |d 1979  |g 46(2024), 5 vom: 03. Apr., Seite 3938-3954  |w (DE-627)NLM098212257  |x 1939-3539  |7 nnns 
773 1 8 |g volume:46  |g year:2024  |g number:5  |g day:03  |g month:04  |g pages:3938-3954 
856 4 0 |u http://dx.doi.org/10.1109/TPAMI.2024.3350380  |3 Volltext 
912 |a GBV_USEFLAG_A 
912 |a SYSFLAG_A 
912 |a GBV_NLM 
912 |a GBV_ILN_350 
951 |a AR 
952 |d 46  |j 2024  |e 5  |b 03  |c 04  |h 3938-3954