Research map / Video / Action recognition
SVFormer: Semi-Supervised Video Transformer for Action Recognition
Read the paperCheck 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 SVFormer-S instead of the strongest listed baseline | Semi-supervised action recognition · Kinetics-400 1% | Top-1 accuracy: 17.6 -> 32.6 | Helped |
| Use SVFormer-S instead of the strongest listed baseline | Semi-supervised action recognition · UCF-101 1% | Top-1 accuracy: 25.1 -> 31.4 | Helped |
| Use SVFormer-S instead of the strongest listed baseline | Semi-supervised action recognition · HMDB-51 40% | Top-1 accuracy: 46.5 -> 56.2 | Helped |
| Replace FixMatch with EMA-Teacher (ablation) | SVFormer-S · UCF-101 1% | Top-1 accuracy: 25.1 -> 31.4 | Helped |
| Replace FixMatch with EMA-Teacher (ablation) | SVFormer-S · Kinetics-400 1% | Top-1 accuracy: 28.2 -> 32.6 | Helped |
| Add Tube TokenMix instead of using CutMix (ablation) | SVFormer-S · UCF-101 1% | Top-1 accuracy: 28.7 -> 31.4 | Helped |
| Add Tube TokenMix instead of using Mixup (ablation) | SVFormer-S · UCF-101 1% | Top-1 accuracy: 29.8 -> 31.4 | Helped |
| Use Tube TokenMix rather than Frame TokenMix (ablation) | SVFormer-S · Kinetics-400 1% | Top-1 accuracy: 26.3 -> 32.6 | Helped |
| Use Tube TokenMix rather than Rand TokenMix (ablation) | SVFormer-S · Kinetics-400 1% | Top-1 accuracy: 28.8 -> 32.6 | Helped |
| Add spatial augmentation to temporal warping (ablation) | SVFormer-S with Tube TokenMix · UCF-101 1% | Top-1 accuracy: 30.2 -> 31.4 | Helped |
| Add temporal warping to spatial augmentation (ablation) | SVFormer-S with Tube TokenMix · Kinetics-400 1% | Top-1 accuracy: 30.3 -> 32.6 | Helped |
| Use video-based sparse sampling (8×32, 1×3) rather than clip-based 8×8, 5×3 (ablation) | SVFormer-S inference · Kinetics-400 1% | Top-1 accuracy: 32.6 -> 31.0 | Hurt |
Closest papers by meaning
- Learning From Temporal Gradient for Semi-Supervised Action Recognition (CVPR 2022 · similarity 0.90)
- Adapting Short-Term Transformers for Action Detection in Untrimmed Videos (CVPR 2024 · similarity 0.87)
- Stable Mean Teacher for Semi-supervised Video Action Detection (AAAI 2025 · similarity 0.87)
- Semi-supervised Active Learning for Video Action Detection (AAAI 2024 · similarity 0.86)
- TrAction: Action Recognition with Sparse Trajectories (NeurIPS 2026 · similarity 0.86)
- SeFAR: Semi-supervised Fine-grained Action Recognition with Temporal Perturbation and Learning Stabilization (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.