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

arXiv:2403.04398v1 Announce Type: new
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. …

abstract arxiv assessment continual cs.cv data diverse feature features graph however inputs manifold memory quality raw refine regularization skills struggle type