Research map / Video / Action recognition
Temporal Action Segmentation From Timestamp Supervision
Check 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 |
|---|---|---|---|
| Train using timestamp-derived frame labels and model-estimated action changes rather than sparse timestamp labels alone | Temporal action segmentation · 50Salads | F1@50: 34.0 -> 60.1 | Helped |
| Add the confidence loss to classification and smoothing losses (ablation) | Timestamp-supervised segmentation · 50Salads | Acc: 74.6 -> 75.6 | Helped |
| Add the confidence loss to classification and smoothing losses (ablation) | Timestamp-supervised segmentation · Breakfast | Acc: 63.7 -> 64.1 | Helped |
| Add the smoothing loss to classification loss (ablation) | Timestamp-supervised segmentation · 50Salads | Acc: 72.8 -> 74.6 | Helped |
| Add the smoothing loss to classification loss (ablation) | Timestamp-supervised segmentation · Breakfast | Acc: 60.2 -> 63.7 | Helped |
| Use forward-backward action-change estimates rather than stamp-to-stamp feature estimates alone (ablation) | Timestamp-supervised segmentation · Breakfast | F1@50: 44.4 -> 47.4 | Helped |
| Use feature-cluster distance energy instead of maximizing segment probabilities (ablation) | Timestamp-supervised segmentation · 50Salads | F1@50: 48.6 -> 60.1 | Helped |
| Use random timestamps rather than timestamps at action start (ablation) | Timestamp-supervised segmentation · Breakfast | F1@50: 28.0 -> 47.4 | Helped |
| Use random timestamps rather than center-frame timestamps (ablation) | Timestamp-supervised segmentation · Breakfast | F1@50: 45.4 -> 47.4 | Helped |
| Increase confidence-loss weight from β = 0 to β = 0.075 (ablation) | Timestamp-supervised segmentation · 50Salads | F1@50: 55.3 -> 60.1 | Helped |
| Use MS-TCN++ rather than MS-TCN as the segmentation model (ablation) | Timestamp-supervised segmentation · 50Salads | F1@50: 57.0; 55.8 | Mixed |
| Use timestamp supervision rather than full frame-wise supervision | Temporal action segmentation · 50Salads | F1@50: 58.6 -> 60.1 | Helped |
Closest papers by meaning
- Reducing the Label Bias for Timestamp Supervised Temporal Action Segmentation (CVPR 2023 · similarity 0.92)
- OnlineTAS: An Online Baseline for Temporal Action Segmentation (NeurIPS 2024 · similarity 0.90)
- FACT: Frame-Action Cross-Attention Temporal Modeling for Efficient Action Segmentation (CVPR 2024 · similarity 0.88)
- Efficient and Effective Weakly-Supervised Action Segmentation via Action-Transition-Aware Boundary Alignment (CVPR 2024 · similarity 0.88)
- Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature (AAAI 2024 · similarity 0.87)
- HR-Pro: Point-Supervised Temporal Action Localization via Hierarchical Reliability Propagation (AAAI 2024 · similarity 0.87)
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.