Research map / Video / Action recognition
What and How Well You Performed? A Multitask Learning Approach to Action Quality Assessment
Paritosh Parmar, Brendan Tran Morris
Check your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add action classification and commentary generation as auxiliary tasks | C3D-AVG AQA on MTL-AQA · MTL-AQA | Spearman's rank correlation: 89.60 -> 90.44 | Improved |
| Add action classification and commentary generation as auxiliary tasks | MSCADC AQA on MTL-AQA · MTL-AQA | Spearman's rank correlation: 84.72 -> 86.12 | Improved |
| Add action classification without commentary generation (ablation) | C3D-AVG AQA · MTL-AQA | Spearman's rank correlation: 89.60 -> 89.62 | No clear effect |
| Add commentary generation without action classification (ablation) | C3D-AVG AQA · MTL-AQA | Spearman's rank correlation: 89.60 -> 88.78 | Hurt |
| Combine action classification with commentary generation (ablation) | C3D-AVG AQA · MTL-AQA | Spearman's rank correlation: 88.78 -> 90.44 | Improved |
| Add action classification without commentary generation (ablation) | MSCADC AQA · MTL-AQA | Spearman's rank correlation: 84.72 -> 85.76 | Improved |
| Add commentary generation without action classification (ablation) | MSCADC AQA · MTL-AQA | Spearman's rank correlation: 84.72 -> 85.47 | Improved |
| Add action classification and commentary generation | C3D-AVG AQA · UNLV Dive | Spearman's rank correlation: 83.83 -> 88.08 | Improved |
| Add action classification and commentary generation | MSCADC AQA · UNLV Dive | Spearman's rank correlation: 79.79 -> 80.60 | Improved |
| Add action classification and commentary generation | C3D-AVG AQA · MTL-AQA with reduced training data | Spearman's rank correlation: 77.27 -> 83.52; 69.63 -> 72.09; 64.17 -> 68.16 | Improved |
| Train AQA-oriented representations jointly with action and commentary supervision, rather than using action-recognition features | C3D-AVG-MTL representations · MTL-AQA convolutional-layer features | linear-regressor Spearman correlation: 74.26 vs. 71.01; 77.95 vs. 71.39; 82.78 vs. 73.13; 86.18 vs. 76.34; 85.75 vs. 73.69 | Improved |
Closest papers by meaning
- A Decade of Action Quality Assessment: Largest Systematic Survey of Trends, Challenges, and Future Directions (IJCV 2026 · similarity 0.90)
- Action Quality Assessment with Temporal Parsing Transformer (ECCV 2022 · similarity 0.89)
- LucidAction: A Hierarchical and Multi-model Dataset for Comprehensive Action Quality Assessment (NeurIPS 2024 · similarity 0.86)
- FineParser: A Fine-grained Spatio-temporal Action Parser for Human-centric Action Quality Assessment (CVPR 2024 · similarity 0.85)
- GAIA: Rethinking Action Quality Assessment for AI-Generated Videos (NeurIPS 2024 · similarity 0.84)
- ActionAtlas: A VideoQA Benchmark for Domain-specialized Action Recognition (NeurIPS 2024 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
