Research map / Video / Action recognition
Representation Learning via Global Temporal Alignment and Cycle-Consistency
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 the contrastive cost (ablation) | SmoothDTW · FineGym101 | accuracy: 28.20 -> 47.32 | Helped |
| Add GCC when contrastive cost is absent (ablation) | SmoothDTW · FineGym101 | accuracy: 28.20 -> 28.20 | No clear effect |
| Add GCC (ablation) | SmoothDTW with contrastive cost · FineGym101 | accuracy: 47.32 -> 49.51 | Helped |
| Replace smoothMin with minγ (ablation) | SmoothDTW with contrastive cost and GCC · FineGym101 | accuracy: 49.51 -> 48.07 | Hurt |
| Use global temporal alignment rather than independently matched frame correspondences | FineGym101, scratch · FineGym101, scratch | accuracy: 41.78 -> 45.79 | Helped |
| Use global temporal alignment rather than independently matched frame correspondences | FineGym290, scratch · FineGym290, scratch | accuracy: 40.57 -> 43.49 | Helped |
| Use global temporal alignment rather than independently matched frame correspondences | FineGym101, only BN · FineGym101, only BN | accuracy: 45.62 -> 49.51 | Helped |
| Use global temporal alignment rather than independently matched frame correspondences | FineGym290, only BN · FineGym290, only BN | accuracy: 43.40 -> 46.54 | Helped |
| Use global temporal alignment rather than independently matched frame correspondences | PennAction joint representation · PennAction, 13 non-repetitive classes | Kendall's Tau: 71.0 -> 74.84 | Helped |
| Use global temporal alignment rather than independently matched frame correspondences | PennAction joint representation · PennAction, 13 non-repetitive classes | Phase classification: 77.51 -> 78.90 | Helped |
Closest papers by meaning
- Temporal Cycle-Consistency Learning (CVPR 2019 · similarity 0.92)
- Learning by Aligning Videos in Time (CVPR 2021 · similarity 0.91)
- D3TW: Discriminative Differentiable Dynamic Time Warping for Weakly Supervised Action Alignment and Segmentation (CVPR 2019 · similarity 0.91)
- Temporal Alignment-Free Video Matching for Few-shot Action Recognition (CVPR 2025 · similarity 0.87)
- Similar Modality Enhancement and Action Consistency Learning for Weakly Supervised Temporal Action Localization (AAAI 2025 · similarity 0.86)
- Temporally Consistent Object-Centric Learning by Contrasting Slots (CVPR 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.