Ego-$A{\mathbf{3}}$: Adaptive Fusion-Based Disentangled Transformer for Egocentric Action Anticipation
Minhyuk Kim, Jong Won Jung, Eungi Lee, Seok Bong Yoo
Abstract
Recently, egocentric action anticipation for wearable robotics cameras has gained considerable attention due to its capability to analyze nouns and verbs from a firstperson view. However, this field encounters challenges due to various uncertainties, such as action-irrelevant information and semantically fused representations of verbs and nouns. To overcome these issues, we introduce Ego- <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$A^{3}$</tex>, designed to improve the robustness and reliability of egocentric action anticipation systems. Ego- <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$A^{3}$</tex> adaptively extracts actionrelevant data to efficiently utilize additional information beyond visual data. Additionally, Ego- <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$A^{3}$</tex> produces effective disentangled representations for verbs and nouns by employing learnable verb and noun queries. Experiments on the EpicKitchens-100 and EGTEA Gaze+ datasets demonstrate that Ego- <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$A^{3}$</tex> outperforms existing methods in top-1 accuracy and mean top- 5 recall. Our code is publicly available at https://github.com/alsgur0720/egocentricanticipation.
BibTeX
@inproceedings{icra2025_egoamathbf3adapt,
title = {Ego-$A{\mathbf{3}}$: Adaptive Fusion-Based Disentangled Transformer for Egocentric Action Anticipation},
author = {Minhyuk Kim and Jong Won Jung and Eungi Lee and Seok Bong Yoo},
booktitle = {ICRA 2025},
year = {2025}
}