Research map / Video / Action recognition
Video Action Segmentation via Contextually Refined Temporal Keypoints
Borui Jiang, Yang Jin, Zhentao Tan, Yadong Mu
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 |
|---|---|---|---|
| replace frame-wise classification with temporal keypoints, refinement, GCN and graph matching (ablation) | ASFormer · GTEA | F1@50: 71.9 -> 83.4 | Improved |
| use PCA graph matching instead of CIE (ablation) | RTK · GTEA | F1@50: 74.5 -> 83.4 | Improved |
| detach graph matching embeddings from keypoint generator gradients (ablation) | RTK · GTEA | F1@50: 78.6 -> 83.4 | Improved |
| increase candidate ratio alpha from 0.001 to 0.1 (ablation) | RTK · GTEA | F1@50: 74.6 -> 83.4 | Improved |
| increase candidate ratio alpha from 0.1 to 1.0 (ablation) | RTK · GTEA | F1@50: 83.4 -> 82.7 | No clear effect |
| RTK versus ASFormer on parameters | RTK · Table 5 comparison | parameters: 1.134M -> 0.779M | Improved |
| use RTK on keypoint accuracy versus frame-wise extraction | ASFormer · 50salads keypoint quality | boundary accuracy: 39.5 -> 92.1 | Improved |
| apply refinement stages after keypoint generator (ablation) | RTK · 50salads keypoint quality | boundary recall: boundary Acc 86.3 -> 92.1, Rec 80.7 -> 71.1 | Mixed |
| replace HASR refinement with RTK | ASFormer · GTEA | F1@50: 79.3 -> 83.4 | Improved |
| add Br-prompt features to RTK | RTK · GTEA | F1@50: 83.0 -> 87.8 | Improved |
Closest papers by meaning
- Temporally-Weighted Hierarchical Clustering for Unsupervised Action Segmentation (CVPR 2021 · similarity 0.92)
- Temporal Action Segmentation: An Analysis of Modern Techniques (TPAMI 2023 · similarity 0.91)
- Unsupervised Action Segmentation via Fast Learning of Semantically Consistent Actoms (AAAI 2024 · similarity 0.91)
- FACT: Frame-Action Cross-Attention Temporal Modeling for Efficient Action Segmentation (CVPR 2024 · similarity 0.90)
- OnlineTAS: An Online Baseline for Temporal Action Segmentation (NeurIPS 2024 · similarity 0.89)
- Efficient and Effective Weakly-Supervised Action Segmentation via Action-Transition-Aware Boundary Alignment (CVPR 2024 · similarity 0.89)
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.
