Disentangling Static and Dynamic Information for Reducing Static Bias in Action Recognition

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Kobayashi, Masato, Ding, Ning, Tamaki, Toru
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911179547344896
author Kobayashi, Masato
Ding, Ning
Tamaki, Toru
author_facet Kobayashi, Masato
Ding, Ning
Tamaki, Toru
contents Action recognition models rely excessively on static cues rather than dynamic human motion, which is known as static bias. This bias leads to poor performance in real-world applications and zero-shot action recognition. In this paper, we propose a method to reduce static bias by separating temporal dynamic information from static scene information. Our approach uses a statistical independence loss between biased and unbiased streams, combined with a scene prediction loss. Our experiments demonstrate that this method effectively reduces static bias and confirm the importance of scene prediction loss.
format Preprint
id arxiv_https___arxiv_org_abs_2509_23009
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Disentangling Static and Dynamic Information for Reducing Static Bias in Action Recognition
Kobayashi, Masato
Ding, Ning
Tamaki, Toru
Computer Vision and Pattern Recognition
Action recognition models rely excessively on static cues rather than dynamic human motion, which is known as static bias. This bias leads to poor performance in real-world applications and zero-shot action recognition. In this paper, we propose a method to reduce static bias by separating temporal dynamic information from static scene information. Our approach uses a statistical independence loss between biased and unbiased streams, combined with a scene prediction loss. Our experiments demonstrate that this method effectively reduces static bias and confirm the importance of scene prediction loss.
title Disentangling Static and Dynamic Information for Reducing Static Bias in Action Recognition
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2509.23009