Contrastive Self-Supervised Learning for Skeleton Action Recognition

Xuehao Gao, Yang Yang, Shaoyi Du
NeurIPS 2020 Workshop on Pre-registration in Machine Learning, PMLR 148:51-61, 2021.

Abstract

Learning discriminative features plays a significant role in action recognition. Many attempts have been made to train deep neural networks by their labeled data. However, in previous networks, the view or distance variations can cause the intra-class differences even larger than inter-class differences. In this work, we propose a new contrastive self-supervised learning method for action recognition of unlabeled skeletal videos. Through contrastive representation learning by adequate compositions of viewpoints and distances, the self-supervised net selects discriminative features which have invariance motion semantics for action recognition. We hope this attempt can be helpful for the unsupervised learning study of skeleton-based action recognition.

Cite this Paper


BibTeX
@InProceedings{pmlr-v148-gao21a, title = {Contrastive Self-Supervised Learning for Skeleton Action Recognition}, author = {Gao, Xuehao and Yang, Yang and Du, Shaoyi}, booktitle = {NeurIPS 2020 Workshop on Pre-registration in Machine Learning}, pages = {51--61}, year = {2021}, editor = {Bertinetto, Luca and Henriques, João F. and Albanie, Samuel and Paganini, Michela and Varol, Gül}, volume = {148}, series = {Proceedings of Machine Learning Research}, month = {11 Dec}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v148/gao21a/gao21a.pdf}, url = {http://proceedings.mlr.press/v148/gao21a.html}, abstract = {Learning discriminative features plays a significant role in action recognition. Many attempts have been made to train deep neural networks by their labeled data. However, in previous networks, the view or distance variations can cause the intra-class differences even larger than inter-class differences. In this work, we propose a new contrastive self-supervised learning method for action recognition of unlabeled skeletal videos. Through contrastive representation learning by adequate compositions of viewpoints and distances, the self-supervised net selects discriminative features which have invariance motion semantics for action recognition. We hope this attempt can be helpful for the unsupervised learning study of skeleton-based action recognition.} }
Endnote
%0 Conference Paper %T Contrastive Self-Supervised Learning for Skeleton Action Recognition %A Xuehao Gao %A Yang Yang %A Shaoyi Du %B NeurIPS 2020 Workshop on Pre-registration in Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Luca Bertinetto %E João F. Henriques %E Samuel Albanie %E Michela Paganini %E Gül Varol %F pmlr-v148-gao21a %I PMLR %P 51--61 %U http://proceedings.mlr.press/v148/gao21a.html %V 148 %X Learning discriminative features plays a significant role in action recognition. Many attempts have been made to train deep neural networks by their labeled data. However, in previous networks, the view or distance variations can cause the intra-class differences even larger than inter-class differences. In this work, we propose a new contrastive self-supervised learning method for action recognition of unlabeled skeletal videos. Through contrastive representation learning by adequate compositions of viewpoints and distances, the self-supervised net selects discriminative features which have invariance motion semantics for action recognition. We hope this attempt can be helpful for the unsupervised learning study of skeleton-based action recognition.
APA
Gao, X., Yang, Y. & Du, S.. (2021). Contrastive Self-Supervised Learning for Skeleton Action Recognition. NeurIPS 2020 Workshop on Pre-registration in Machine Learning, in Proceedings of Machine Learning Research 148:51-61 Available from http://proceedings.mlr.press/v148/gao21a.html.

Related Material