Research map / Label-efficient and robust learning / Semi-supervised learning
Semi-Supervised Teacher-Reference-Student Architecture for Action Quality Assessment
Wulian Yun, Mengshi Qi, Fei Peng, Huadong Ma
Read the paperCheck 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 reference network to teacher-student baseline (ablation) | teacher-student baseline · MTL-AQA (10% labels) | Spearman correlation: 0.751 -> 0.811 | Improved |
| add reference network to teacher-student baseline (ablation) | teacher-student baseline · MTL-AQA (40% labels) | Spearman correlation: 0.865 -> 0.891 | Improved |
| add confidence memory to teacher network (ablation) | teacher-student baseline · MTL-AQA (10% labels) | Spearman correlation: 0.751 -> 0.794 | Improved |
| add confidence memory to teacher network (ablation) | teacher-student baseline · MTL-AQA (40% labels) | Spearman correlation: 0.865 -> 0.870 | No clear effect |
| update memory by confidence instead of direct current-output update (ablation) | Base + RN · MTL-AQA (10% labels) | Spearman correlation: 0.811 -> 0.813 | Improved |
| add confidence memory to reference network (ablation) | Base + RN + TM · MTL-AQA (10% labels) | Spearman correlation: 0.813 -> 0.825 | Improved |
| add confidence memory to reference network (ablation) | Base + RN + TM · MTL-AQA (40% labels) | Spearman correlation: 0.893 -> 0.901 | Improved |
| increase labeled data proportion from 10% to 40% (ablation) | Ours · MTL-AQA | Spearman correlation: 0.825 -> 0.901 | Improved |
| use semi-supervised teacher-reference-student instead of S4AQA | Ours · Rhythmic Gymnastics (40% labels, average) | Spearman correlation: 0.342 -> 0.529 | Improved |
| use semi-supervised teacher-reference-student instead of S4AQA | Ours · JIGSAWS Knot Tying (50% labels) | Spearman correlation: 0.813 -> 0.753 | Hurt |
Closest papers by meaning
- Group-Aware Contrastive Regression for Action Quality Assessment (ICCV 2021 · similarity 0.86)
- Action Quality Assessment with Temporal Parsing Transformer (ECCV 2022 · similarity 0.86)
- A Decade of Action Quality Assessment: Largest Systematic Survey of Trends, Challenges, and Future Directions (IJCV 2026 · similarity 0.85)
- MoAKE: Toward Unified All-in-One Action Quality Assessment via Mixture of Action Knowledge Experts (ECCV 2026 · similarity 0.85)
- LucidAction: A Hierarchical and Multi-model Dataset for Comprehensive Action Quality Assessment (NeurIPS 2024 · similarity 0.84)
- Stable Mean Teacher for Semi-supervised Video Action Detection (AAAI 2025 · similarity 0.84)
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.
