Researchers have developed DualAlign, a novel two-stage framework designed to improve Action Quality Assessment (AQA) by effectively fusing multi-modal data. This approach addresses challenges like cross-modal misalignment and the high cost of annotation by first creating a stable visual representation from RGB, optical flow, and skeleton data, and then integrating textual semantics. To test DualAlign, a new dataset called MM--JDM was introduced, which includes noisy and imbalanced multi-modal data. Experiments demonstrate that DualAlign significantly outperforms existing methods on MM--JDM and other benchmarks, even under conditions with missing modalities or scarce labels. AI
IMPACT This research could improve automated scoring in sports, skill assessment, and healthcare by enabling more accurate evaluation of human movement quality.
RANK_REASON The cluster contains a research paper detailing a new framework and dataset for action quality assessment. [lever_c_demoted from research: ic=1 ai=1.0]
- Action Quality Assessment
- alphaXiv
- arXiv
- CatalyzeX
- DagsHub
- Gotit.pub
- Hugging Face
- MM--JDM
- ScienceCast
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
