Research map / Video / Action recognition
Cross-Modal Learning with 3D Deformable Attention for Action Recognition
Sangwon Kim, Dasom Ahn, Byoung Chul Ko
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 3D deformable attention (ablation) | 3D deformable transformer · PennAction | top-1 accuracy (%): 94.8 -> 99.7 | Improved |
| add joint stride attention (ablation) | 3D deformable transformer · PennAction | top-1 accuracy (%): 98.2 -> 99.7 | Improved |
| add temporal stride attention (ablation) | 3D deformable transformer · PennAction | top-1 accuracy (%): 98.5 -> 99.7 | Improved |
| use cross-modal tokens instead of a single shared token (ablation) | 3D deformable transformer · PennAction | top-1 accuracy (%): 97.0 -> 99.7 | Improved |
| use cross-modal tokens instead of averaging RGB and pose tokens (ablation) | 3D deformable transformer · PennAction | top-1 accuracy (%): 92.2 -> 99.7 | Improved |
| set temporal stride to about half the window size (ablation) | temporal stride attention · PennAction | top-1 accuracy | Improved |
| use 12 input frames instead of fewer (ablation) | 3D deformable transformer · PennAction | top-1 accuracy | Improved |
| train without pre-training versus pre-trained PoseC3D | 3D deformable transformer · FineGYM | mean accuracy (%): 95.6 -> 90.3 | Hurt |
Closest papers by meaning
- Deformable Video Transformer (CVPR 2022 · similarity 0.88)
- Interact Before Align: Leveraging Cross-Modal Knowledge for Domain Adaptive Action Recognition (CVPR 2022 · similarity 0.87)
- Shrinking Temporal Attention in Transformers for Video Action Recognition (AAAI 2022 · similarity 0.86)
- CMD: Self-Supervised 3D Action Representation Learning with Cross-Modal Mutual Distillation (ECCV 2022 · similarity 0.86)
- Global Context-Aware Attention LSTM Networks for 3D Action Recognition (CVPR 2017 · similarity 0.86)
- Toward Stronger 3D Human Action Representation Learning Via Efficient Spatiotemporal Decoupling (TPAMI 2026 · 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.
