Research map / Video / Action recognition
Unified Fully and Timestamp Supervised Temporal Action Segmentation via Sequence to Sequence Translation
Nadine Behrmann, S. Alireza Golestaneh, Zico Kolter, Jürgen Gall, Mehdi Noroozi
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 frame-wise loss on the encoder to the vanilla Transformer (ablation) | UVAST · Breakfast split 1 | Edit: 52.9 -> 76.5 | Improved |
| add alignment decoder for implicit duration instead of explicit MSE duration (ablation) | UVAST · Breakfast split 1 | Acc: 59.0 -> 71.7 | Improved |
| add cross-attention loss for the transcript decoder (ablation) | UVAST (L_frame + L_segment) · 50Salads split 1 | Edit: 45.3 -> 65.5 | Improved |
| add cross-attention loss for the transcript decoder (ablation) | UVAST (L_frame + L_segment) · Breakfast split 1 | Edit: 73.4 -> 74.5 | Improved |
| add frame-wise loss to the segment-wise loss (ablation) | UVAST · Breakfast split 1 | Edit: 55.6 -> 73.4 | Improved |
| add group-wise losses on top of cross-attention loss (ablation) | UVAST · 50Salads split 1 | Edit: 65.5 -> 78.4 | Improved |
| add both group-wise losses without cross-attention loss (ablation) | UVAST (L_frame + L_segment) · 50Salads split 1 | Edit: 45.3 -> 42.4 | Hurt |
| use proposed encoder instead of ASFormer encoder (ablation) | UVAST · Breakfast split 1 | Edit: 74.6 -> 76.1 | Improved |
| add split-segment training (ablation) | UVAST · GTEA split 1 | Edit: 88.2 -> 93.4 | Improved |
| change split-segment value from 0.05 to 0.17 (ablation) | UVAST · Breakfast split 1 | Edit: 74.9 -> 76.1 | Improved |
| use temporal continuity constraint in k-medoids (ablation) | k-medoids pseudo-segmentation · Breakfast | Edit: 12.3 -> 100.0 | Improved |
| add Viterbi post-processing | UVAST · 50Salads fully supervised | Acc: 79.5 -> 87.4 | Improved |
Closest papers by meaning
- Temporal Action Segmentation From Timestamp Supervision (CVPR 2021 · similarity 0.90)
- Efficient and Effective Weakly-Supervised Action Segmentation via Action-Transition-Aware Boundary Alignment (CVPR 2024 · similarity 0.88)
- Seq2Time: Sequential Knowledge Transfer for Video LLM Temporal Grounding (CVPR 2025 · similarity 0.88)
- FACT: Frame-Action Cross-Attention Temporal Modeling for Efficient Action Segmentation (CVPR 2024 · similarity 0.88)
- OnlineTAS: An Online Baseline for Temporal Action Segmentation (NeurIPS 2024 · similarity 0.86)
- Hierarchical Vector Quantization for Unsupervised Action Segmentation (AAAI 2025 · similarity 0.86)
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.
