MAGR: Manifold-Aligned Graph Regularization for Continual Action Quality Assessment

Kanglei Zhou, Liyuan Wang, Xingxing Zhang, Hubert P. H. Shum, Frederick W. B. Li, Jianguo Li and Xiaohui Liang
Proceedings of the 2024 European Conference on Computer Vision (ECCV), 2024

Oral Paper (Top 2.3% of 8585 Submissions) H5-Index: 206# Core A* Conference

MAGR: Manifold-Aligned Graph Regularization for Continual Action Quality Assessment
‡ According to Core Ranking 2023
# According to Google Scholar 2024

Abstract

Action Quality Assessment (AQA) evaluates diverse skills but models struggle with non-stationary data. We propose Continual AQA (CAQA) to refine models using sparse new data. Feature replay preserves memory without storing raw inputs. However, the misalignment between static old features and the dynamically changing feature manifold causes severe catastrophic forgetting. To address this novel problem, we propose Manifold-Aligned Graph Regularization (MAGR), which first aligns deviated old features to the current feature manifold, ensuring representation consistency. It then constructs a graph jointly arranging old and new features aligned with quality scores. Experiments show MAGR outperforms recent strong baselines with up to 6.56%, 5.66%, 15.64%, and 9.05% correlation gains on the MTL-AQA, FineDiving, UNLV-Dive, and JDM-MSA split datasets, respectively. This validates MAGR for continual assessment challenges arising from non-stationary skill variation.


Downloads


YouTube


Cite This Research

Plain Text

Kanglei Zhou, Liyuan Wang, Xingxing Zhang, Hubert P. H. Shum, Frederick W. B. Li, Jianguo Li and Xiaohui Liang, "MAGR: Manifold-Aligned Graph Regularization for Continual Action Quality Assessment," in ECCV '24: Proceedings of the 2024 European Conference on Computer Vision, vol. 15069, pp. 375-392, Milan, Italy, Springer, 2024.

BibTeX

@inproceedings{zhou24magr,
 author={Zhou, Kanglei and Wang, Liyuan and Zhang, Xingxing and Shum, Hubert P. H. and Li, Frederick W. B. and Li, Jianguo and Liang, Xiaohui},
 booktitle={Proceedings of the 2024 European Conference on Computer Vision},
 series={ECCV '24},
 title={MAGR: Manifold-Aligned Graph Regularization for Continual Action Quality Assessment},
 year={2024},
 volume={15069},
 pages={375--392},
 numpages={18},
 doi={10.1007/978-3-031-73247-8_22},
 publisher={Springer},
 location={Milan, Italy},
}

RIS

TY  - CONF
AU  - Zhou, Kanglei
AU  - Wang, Liyuan
AU  - Zhang, Xingxing
AU  - Shum, Hubert P. H.
AU  - Li, Frederick W. B.
AU  - Li, Jianguo
AU  - Liang, Xiaohui
T2  - Proceedings of the 2024 European Conference on Computer Vision
TI  - MAGR: Manifold-Aligned Graph Regularization for Continual Action Quality Assessment
PY  - 2024
VL  - 15069
SP  - 375
EP  - 392
DO  - 10.1007/978-3-031-73247-8_22
PB  - Springer
ER  - 


Supporting Grants


Similar Research

Kanglei Zhou, Yue Ma, Hubert P. H. Shum and Xiaohui Liang, "Hierarchical Graph Convolutional Networks for Action Quality Assessment", IEEE Transactions on Circuits and Systems for Video Technology (TCSVT), 2023
Kanglei Zhou, Ruizhi Cai, Yue Ma, Qingqing Tan, Xinning Wang, Jianguo Li, Hubert P. H. Shum, Frederick W. B. Li, Song Jin and Xiaohui Liang, "A Video-Based Augmented Reality System for Human-in-the-Loop Muscle Strength Assessment of Juvenile Dermatomyositis", IEEE Transactions on Visualization and Computer Graphics (TVCG) - Proceedings of the 2023 IEEE Conference on Virtual Reality and 3D User Interfaces (VR), 2023

 

Last updated on 17 November 2024
RSS Feed