Research map / Video / Action recognition
SCT: Set Constrained Temporal Transformer for Set Supervised Action Segmentation
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 |
|---|---|---|---|
| Add region loss to the set-loss configuration (ablation) | SCT · Breakfast split 1 | MoF: 8.1 -> 9.9 | Helped |
| Add inverse sparsity regularization (ablation) | SCT · Breakfast split 1 | MoF: 9.9 -> 19.2 | Helped |
| Add temporal consistency loss (ablation) | SCT · Breakfast split 1 | MoF: 19.2 -> 21.9 | Helped |
| Add SCT self-supervision loss to train region lengths (ablation) | SCT · Breakfast split 1 | MoF: 21.9 -> 29.9 | Helped |
| Add region length regularization to the configuration with self-supervision (ablation) | SCT · Breakfast split 1 | MoF: 29.9 -> 30.8 | Helped |
| Use the length regularizer without SCT self-supervision (ablation) | SCT · Breakfast split 1 | MoF: 21.9 -> 22.2 | Hurt |
| Replace SCT self-supervision with Jensen-Shannon divergence (ablation) | SCT · Breakfast split 1 | MoF: 29.9 -> 25.3 | Hurt |
| Use 3 max-pooling operations rather than 0 (ablation) | SCT temporal embedding · Breakfast split 1 | MoF: 12.3 -> 28.1 | Helped |
| Increase max-pooling operations from 3 to 6 (ablation) | SCT temporal embedding · Breakfast split 1 | MoF: 28.1 -> 18.2 | Hurt |
| Use SCT instead of ActionSet-monte-carlo | Set-supervised action segmentation · Breakfast | MoF: 23.3 -> 30.4 | Helped |
| Use SCT instead of ActionSet-monte-carlo | Set-supervised action segmentation · Hollywood Extended | Jaccard index: 9.3 -> 17.7 | Helped |
| Use SCT instead of ActionSet-monte-carlo | Set-supervised action segmentation · Cooking 2 | midpoint hit: 9.8 -> 14.3 | Helped |
Closest papers by meaning
- Action Sets: Weakly Supervised Action Segmentation Without Ordering Constraints (CVPR 2018 · similarity 0.94)
- Iterative Contrast-Classify for Semi-supervised Temporal Action Segmentation (AAAI 2022 · similarity 0.91)
- Weakly-Supervised Action Segmentation With Iterative Soft Boundary Assignment (CVPR 2018 · similarity 0.90)
- FACT: Frame-Action Cross-Attention Temporal Modeling for Efficient Action Segmentation (CVPR 2024 · similarity 0.89)
- Efficient and Effective Weakly-Supervised Action Segmentation via Action-Transition-Aware Boundary Alignment (CVPR 2024 · similarity 0.89)
- Weakly-Supervised Temporal Action Localization by Inferring Salient Snippet-Feature (AAAI 2024 · similarity 0.88)
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.