Disentangling Static and Dynamic Information for Reducing Static Bias in Action Recognition
Not on the map. Why?
Abstract (arXiv)
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.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 20 of its 30 references that are in the map. First posted on arXiv 27 Sep 2025.
Builds on
20 of 30 references.
References · 20
- 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
- 2015Domain-Adversarial Training of Neural Networkscited by 3,044
- 2017The Kinetics Human Action Video Datasetcited by 1,909
- 2021Is Space-Time Attention All You Need for Video Understanding?cited by 1,620
- 2023MVBench: A Comprehensive Multi-modal Video Understanding Benchmarkcited by 1,109
- 2017The "something something" video database for learning and evaluating visual common sensecited by 1,103
- 2022Revealing Single Frame Bias for Video-and-Language Learningcited by 133
- 2019Why Can't I Dance in the Mall? Learning to Mitigate Scene Bias in Action Recognitioncited by 117
- 2021Evidential Deep Learning for Open Set Action Recognitioncited by 96
- 2019Only Time Can Tell: Discovering Temporal Data for Temporal Modelingcited by 67
- 2020Removing the Background by Adding the Background: Towards Background Robust Self-supervised Video Representation Learningcited by 66
- 2019Semantic-Aware Scene Recognitioncited by 43
8 more
- 2022Mitigating and Evaluating Static Bias of Action Representations in the Background and the Foregroundcited by 27
- 2022Time Is MattEr: Temporal Self-supervision for Video Transformerscited by 11
- 2022ObjectMix: Data Augmentation by Copy-Pasting Objects in Videos for Action Recognitioncited by 7
- 2023DEVIAS: Learning Disentangled Video Representations of Action and Scenecited by 7
- 2025Can masking background and object reduce static bias for zero-shot action recognition?cited by 6
- 2023Learning Human Action Recognition Representations Without Real Humanscited by 5
- 2025ALBAR: Adversarial Learning approach to mitigate Biases in Action Recognitioncited by 5
- 2023S3Aug: Segmentation, Sampling, and Shift for Action Recognitioncited by 3
Built on by
Not cited yet.