Research map / Video / Action recognition
Self-Supervised Video Transformer
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 |
|---|---|---|---|
| Use local-to-global plus global-to-global view correspondences (ablation) | SVT · UCF-101 and HMDB-51 alternate train splits | accuracy: UCF-101: 81.95 -> 84.64; HMDB-51: 49.04 -> 52.17 | Helped |
| Add local-to-local correspondence (ablation) | SVT with local-to-global and global-to-global correspondences · UCF-101 and HMDB-51 alternate train splits | accuracy: UCF-101: 84.64 -> 83.11; HMDB-51: 52.17 -> 51.23 | Hurt |
| Add global-to-local correspondence (ablation) | SVT with local-to-global and global-to-global correspondences · HMDB-51 alternate train split | accuracy: 52.17 -> 51.88 | Hurt |
| Add global-to-local correspondence (ablation) | SVT with local-to-global and global-to-global correspondences · UCF-101 alternate train split | accuracy: 84.64 -> 84.71 | Helped |
| Vary spatial and temporal fields of view together rather than spatially only (ablation) | SVT · UCF-101 and HMDB-51 alternate train splits | accuracy: UCF-101: 73.81 -> 84.64; HMDB-51: 42.91 -> 52.17 | Helped |
| Add temporal variation to spatial variation (ablation) | SVT · UCF-101 and HMDB-51 alternate train splits | accuracy: UCF-101: 73.81 -> 84.64; HMDB-51: 42.91 -> 52.17 | Mixed |
| Use motion correspondence sampling instead of temporal interval sampling (ablation) | SVT · UCF-101 and HMDB-51 alternate train splits | accuracy: UCF-101: 82.24 -> 84.64; HMDB-51: 50.10 -> 52.17 | Helped |
| Add temporally consistent augmentations (ablation) | SVT · UCF-101 and HMDB-51 alternate train splits | accuracy: UCF-101: 84.20 -> 84.64; HMDB-51: 52.10 -> 52.17 | Helped |
| Add slow-fast multi-view inference (ablation) | SVT · UCF-101 and HMDB-51 alternate train splits | accuracy: UCF-101: 84.64 -> 84.80; HMDB-51: 52.17 -> 53.22 | Helped |
Closest papers by meaning
- Time Is MattEr: Temporal Self-supervision for Video Transformers (ICML 2022 · similarity 0.91)
- Long-Short Temporal Contrastive Learning of Video Transformers (CVPR 2022 · similarity 0.90)
- Adapting Short-Term Transformers for Action Detection in Untrimmed Videos (CVPR 2024 · similarity 0.89)
- SVFormer: Semi-Supervised Video Transformer for Action Recognition (CVPR 2023 · similarity 0.89)
- Understanding Video Transformers via Universal Concept Discovery (CVPR 2024 · similarity 0.87)
- No More Shortcuts: Realizing the Potential of Temporal Self-Supervision (AAAI 2024 · 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.