Research map / Label-efficient and robust learning / Semi-supervised learning
Reducing the Label Bias for Timestamp Supervised Temporal Action Segmentation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 MTP before Naive initialization (ablation) | D-TSTAS on GTEA · GTEA | F1@10: 76.0 -> 86.2 | Helped |
| Add MTP before Naive initialization (ablation) | D-TSTAS on 50Salads · 50Salads | F1@10: 72.0 -> 74.5 | Helped |
| Predict timestamp labels from masked features instead of reconstructing input features (ablation) | D-TSTAS · GTEA | F1@10: 75.6 -> 86.2 | Helped |
| Predict timestamp labels from masked features instead of reconstructing input features (ablation) | D-TSTAS · 50Salads | F1@10: 72.3 -> 74.5 | Helped |
| Add CTE refinement (ablation) | D-TSTAS on 50Salads · 50Salads | F1@10: 74.5 -> 76.4 | Helped |
| Increase CTE training epochs per segmentation step from K=10 to K=30 (ablation) | D-TSTAS on 50Salads · 50Salads | F1@10: 80.1 -> 84.4 | Helped |
| Increase CTE training epochs per segmentation step from K=30 to K=40 (ablation) | D-TSTAS on 50Salads · 50Salads | F1@10: 84.4 -> 83.5 | Hurt |
| Increase CTE iterations from M=2 to M=6 (ablation) | D-TSTAS on 50Salads · 50Salads | F1@10: 81.0 -> 84.4 | Helped |
| Increase CTE iterations from M=6 to M=10 (ablation) | D-TSTAS on 50Salads · 50Salads | F1@10: 84.4 -> 83.7 | Hurt |
| Replace confidence loss with segmental confidence loss (ablation) | D-TSTAS on 50Salads · 50Salads | F1@10: 82.9 -> 84.2 | Helped |
| Replace cross-entropy with weighted cross-entropy (ablation) | D-TSTAS on 50Salads · 50Salads | F1@10: 82.5 -> 83.8 | Helped |
| Use D-TSTAS instead of the strongest timestamp-supervised baseline | Temporal action segmentation · GTEA | F1@10: 87.2 -> 91.5 | Helped |
Closest papers by meaning
- Temporal Action Segmentation From Timestamp Supervision (CVPR 2021 · similarity 0.92)
- OnlineTAS: An Online Baseline for Temporal Action Segmentation (NeurIPS 2024 · similarity 0.88)
- FACT: Frame-Action Cross-Attention Temporal Modeling for Efficient Action Segmentation (CVPR 2024 · similarity 0.86)
- HyrCap: Hybrid Rank-Calibration of Action Proposals for Temporal Event Understanding (NeurIPS 2026 · similarity 0.86)
- HR-Pro: Point-Supervised Temporal Action Localization via Hierarchical Reliability Propagation (AAAI 2024 · similarity 0.85)
- Efficient and Effective Weakly-Supervised Action Segmentation via Action-Transition-Aware Boundary Alignment (CVPR 2024 · similarity 0.85)
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.