Seeing Beyond the Scene: Analyzing and Mitigating Background Bias in Action Recognition

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhou, Ellie, Chung, Jihoon, Russakovsky, Olga
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917159472463872
author Zhou, Ellie
Chung, Jihoon
Russakovsky, Olga
author_facet Zhou, Ellie
Chung, Jihoon
Russakovsky, Olga
contents Human action recognition models often rely on background cues rather than human movement and pose to make predictions, a behavior known as background bias. We present a systematic analysis of background bias across classification models, contrastive text-image pretrained models, and Video Large Language Models (VLLM) and find that all exhibit a strong tendency to default to background reasoning. Next, we propose mitigation strategies for classification models and show that incorporating segmented human input effectively decreases background bias by 3.78%. Finally, we explore manual and automated prompt tuning for VLLMs, demonstrating that prompt design can steer predictions towards human-focused reasoning by 9.85%.
format Preprint
id arxiv_https___arxiv_org_abs_2512_17953
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Seeing Beyond the Scene: Analyzing and Mitigating Background Bias in Action Recognition
Zhou, Ellie
Chung, Jihoon
Russakovsky, Olga
Computer Vision and Pattern Recognition
Artificial Intelligence
Human action recognition models often rely on background cues rather than human movement and pose to make predictions, a behavior known as background bias. We present a systematic analysis of background bias across classification models, contrastive text-image pretrained models, and Video Large Language Models (VLLM) and find that all exhibit a strong tendency to default to background reasoning. Next, we propose mitigation strategies for classification models and show that incorporating segmented human input effectively decreases background bias by 3.78%. Finally, we explore manual and automated prompt tuning for VLLMs, demonstrating that prompt design can steer predictions towards human-focused reasoning by 9.85%.
title Seeing Beyond the Scene: Analyzing and Mitigating Background Bias in Action Recognition
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2512.17953