Balanced Representation Learning for Long-tailed Skeleton-based Action Recognition

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Liu, Hongda, Wang, Yunlong, Ren, Min, Hu, Junxing, Luo, Zhengquan, Hou, Guangqi, Sun, Zhenan
Natura: Preprint
Pubblicazione: 2023
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909503190990848
author Liu, Hongda
Wang, Yunlong
Ren, Min
Hu, Junxing
Luo, Zhengquan
Hou, Guangqi
Sun, Zhenan
author_facet Liu, Hongda
Wang, Yunlong
Ren, Min
Hu, Junxing
Luo, Zhengquan
Hou, Guangqi
Sun, Zhenan
contents Skeleton-based action recognition has recently made significant progress. However, data imbalance is still a great challenge in real-world scenarios. The performance of current action recognition algorithms declines sharply when training data suffers from heavy class imbalance. The imbalanced data actually degrades the representations learned by these methods and becomes the bottleneck for action recognition. How to learn unbiased representations from imbalanced action data is the key to long-tailed action recognition. In this paper, we propose a novel balanced representation learning method to address the long-tailed problem in action recognition. Firstly, a spatial-temporal action exploration strategy is presented to expand the sample space effectively, generating more valuable samples in a rebalanced manner. Secondly, we design a detached action-aware learning schedule to further mitigate the bias in the representation space. The schedule detaches the representation learning of tail classes from training and proposes an action-aware loss to impose more effective constraints. Additionally, a skip-modal representation is proposed to provide complementary structural information. The proposed method is validated on four skeleton datasets, NTU RGB+D 60, NTU RGB+D 120, NW-UCLA, and Kinetics. It not only achieves consistently large improvement compared to the state-of-the-art (SOTA) methods, but also demonstrates a superior generalization capacity through extensive experiments. Our code is available at https://github.com/firework8/BRL.
format Preprint
id arxiv_https___arxiv_org_abs_2308_14024
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Balanced Representation Learning for Long-tailed Skeleton-based Action Recognition
Liu, Hongda
Wang, Yunlong
Ren, Min
Hu, Junxing
Luo, Zhengquan
Hou, Guangqi
Sun, Zhenan
Computer Vision and Pattern Recognition
Skeleton-based action recognition has recently made significant progress. However, data imbalance is still a great challenge in real-world scenarios. The performance of current action recognition algorithms declines sharply when training data suffers from heavy class imbalance. The imbalanced data actually degrades the representations learned by these methods and becomes the bottleneck for action recognition. How to learn unbiased representations from imbalanced action data is the key to long-tailed action recognition. In this paper, we propose a novel balanced representation learning method to address the long-tailed problem in action recognition. Firstly, a spatial-temporal action exploration strategy is presented to expand the sample space effectively, generating more valuable samples in a rebalanced manner. Secondly, we design a detached action-aware learning schedule to further mitigate the bias in the representation space. The schedule detaches the representation learning of tail classes from training and proposes an action-aware loss to impose more effective constraints. Additionally, a skip-modal representation is proposed to provide complementary structural information. The proposed method is validated on four skeleton datasets, NTU RGB+D 60, NTU RGB+D 120, NW-UCLA, and Kinetics. It not only achieves consistently large improvement compared to the state-of-the-art (SOTA) methods, but also demonstrates a superior generalization capacity through extensive experiments. Our code is available at https://github.com/firework8/BRL.
title Balanced Representation Learning for Long-tailed Skeleton-based Action Recognition
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2308.14024