Disentangled Concepts Speak Louder Than Words: Explainable Video Action Recognition

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Lee, Jongseo, Lee, Wooil, Park, Gyeong-Moon, Kim, Seong Tae, Choi, Jinwoo
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866918213106794496
author Lee, Jongseo
Lee, Wooil
Park, Gyeong-Moon
Kim, Seong Tae
Choi, Jinwoo
author_facet Lee, Jongseo
Lee, Wooil
Park, Gyeong-Moon
Kim, Seong Tae
Choi, Jinwoo
contents Effective explanations of video action recognition models should disentangle how movements unfold over time from the surrounding spatial context. However, existing methods based on saliency produce entangled explanations, making it unclear whether predictions rely on motion or spatial context. Language-based approaches offer structure but often fail to explain motions due to their tacit nature -- intuitively understood but difficult to verbalize. To address these challenges, we propose Disentangled Action aNd Context concept-based Explainable (DANCE) video action recognition, a framework that predicts actions through disentangled concept types: motion dynamics, objects, and scenes. We define motion dynamics concepts as human pose sequences. We employ a large language model to automatically extract object and scene concepts. Built on an ante-hoc concept bottleneck design, DANCE enforces prediction through these concepts. Experiments on four datasets -- KTH, Penn Action, HAA500, and UCF-101 -- demonstrate that DANCE significantly improves explanation clarity with competitive performance. We validate the superior interpretability of DANCE through a user study. Experimental results also show that DANCE is beneficial for model debugging, editing, and failure analysis.
format Preprint
id arxiv_https___arxiv_org_abs_2511_03725
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Disentangled Concepts Speak Louder Than Words: Explainable Video Action Recognition
Lee, Jongseo
Lee, Wooil
Park, Gyeong-Moon
Kim, Seong Tae
Choi, Jinwoo
Computer Vision and Pattern Recognition
Effective explanations of video action recognition models should disentangle how movements unfold over time from the surrounding spatial context. However, existing methods based on saliency produce entangled explanations, making it unclear whether predictions rely on motion or spatial context. Language-based approaches offer structure but often fail to explain motions due to their tacit nature -- intuitively understood but difficult to verbalize. To address these challenges, we propose Disentangled Action aNd Context concept-based Explainable (DANCE) video action recognition, a framework that predicts actions through disentangled concept types: motion dynamics, objects, and scenes. We define motion dynamics concepts as human pose sequences. We employ a large language model to automatically extract object and scene concepts. Built on an ante-hoc concept bottleneck design, DANCE enforces prediction through these concepts. Experiments on four datasets -- KTH, Penn Action, HAA500, and UCF-101 -- demonstrate that DANCE significantly improves explanation clarity with competitive performance. We validate the superior interpretability of DANCE through a user study. Experimental results also show that DANCE is beneficial for model debugging, editing, and failure analysis.
title Disentangled Concepts Speak Louder Than Words: Explainable Video Action Recognition
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.03725