Research map / Video / Action recognition
Punching Bag vs. Punching Person: Motion Transferability in Videos
Raiyaan Abdullah, Jared Claypoole, Michael Cogswell, Ajay Divakaran, Yogesh Rawat
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 realistic backgrounds with solid plain backgrounds in Syn-TA (ablation) | EZ-CLIP · Syn-TA coarse motions | harmonic mean: 68.38 -> 77.17 | Improved |
| replace realistic backgrounds with solid plain backgrounds in Syn-TA (ablation) | EZ-CLIP · Syn-TA fine motions | harmonic mean: 38.71 -> 51.01 | Improved |
| plain background for all 8 unimodal models (ablation) | unimodal models · Syn-TA coarse motions | harmonic mean: 41.30 -> 51.62 (ResNet50) | Improved |
| add coarse/fine disentanglement branches with residual fine-to-coarse links | EZ-CLIP · Syn-TA coarse | HM: 67.96 -> 70.69 (Set 1), 68.80 -> 72.03 (Set 2) | Improved |
| add coarse/fine disentanglement branches | EZ-CLIP · SSv2-TA coarse | HM: 44.83 -> 46.89 (Set 1), 44.61 -> 47.82 (Set 2) | Improved |
| add coarse/fine disentanglement branches | EZ-CLIP · K400-TA coarse | HM: 76.07 -> 76.06 (Set 1), 74.80 -> 75.20 (Set 2) | No clear effect |
| move from known to unknown fine contexts | all 13 models · Syn-TA, K400-TA, SSv2-TA coarse | accuracy drop: AIM Syn-TA 99.13 -> 70.16 | Hurt |
| increase model size | action recognition models · SSv2-TA | coarse accuracy | No clear effect |
| increase model size | action recognition models · K400-TA | coarse accuracy | Improved |
| use fine captions instead of coarse for unknown classes | EZ-CLIP · K400-TA unknown | accuracy: <30% coarse -> >70% fine | Improved |
| evaluate Syn-TA instead of K400-TA | 11 of 13 models · coarse harmonic mean | harmonic mean | Hurt |
Closest papers by meaning
- About Time: Advances, Challenges, and Outlooks of Action Understanding (IJCV 2025 · similarity 0.88)
- How Transferable are Video Representations Based on Synthetic Data? (NeurIPS 2022 · similarity 0.88)
- Disentangled Concepts Speak Louder Than Words: Explainable Video Action Recognition (NeurIPS 2025 · similarity 0.87)
- ActionAtlas: A VideoQA Benchmark for Domain-specialized Action Recognition (NeurIPS 2024 · similarity 0.86)
- DisMo: Disentangled Motion Representations for Open-World Motion Transfer (NeurIPS 2025 · similarity 0.86)
- FACT: Frame-Action Cross-Attention Temporal Modeling for Efficient Action Segmentation (CVPR 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.
