Research map / Video / Action recognition
Learning Video Representations of Human Motion From Synthetic Data
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 |
|---|---|---|---|
| Pretrain SlowOnly with GATA rather than Kinetics-400 | SlowOnly · Charades-Motion | mAP: 24.5 -> 28.2 | Helped |
| Pretrain SlowOnly with GATA rather than Kinetics-400 | SlowOnly · NTU-RGB X-setup | score: 40.5 -> 53.4 | Helped |
| Pretrain SlowOnly with GATA rather than Kinetics-400 | SlowOnly · Charades | mAP: 35.5 -> 32.0 | Hurt |
| Pretrain SlowOnly with GATA rather than Kinetics-400 | SlowOnly · HAA500 | score: 23.4 -> 21.7 | Hurt |
| Jointly pretrain on GATA, HAA500, and Kinetics-Tracklet rather than GATA alone (ablation) | TSM · Charades | mAP: 31.2 -> 35.7 | Helped |
| Add cross-domain positive-pair mining (ablation) | TSM joint training · Charades | mAP: 35.2 -> 35.7 | Helped |
| Add cross-domain positive-pair mining (ablation) | TSM joint training · GATA | KNN accuracy: 76.5 -> 74.2 | Hurt |
| Replace GATA contrastive training with cross-entropy training (ablation) | SlowOnly · Charades-Motion | mAP: 28.2 -> 27.9 | Hurt |
| Use SlowOnly 101 rather than SlowOnly 50 for joint training (ablation) | SlowOnly · NTU-RGB X-setup | score: 60.7 -> 65.3 | Helped |
| Add HAA500 to GATA pretraining (ablation) | TSM · Charades | mAP: 31.2 -> 34.5 | Helped |
Closest papers by meaning
- Procedural Generation of Videos to Train Deep Action Recognition Networks (CVPR 2017 · similarity 0.88)
- Scaling Large Motion Models with Million-Level Human Motions (ICML 2025 · similarity 0.84)
- DisMo: Disentangled Motion Representations for Open-World Motion Transfer (NeurIPS 2025 · similarity 0.84)
- HumanVid: Demystifying Training Data for Camera-controllable Human Image Animation (NeurIPS 2024 · similarity 0.84)
- MotionRAG: Motion Retrieval-Augmented Image-to-Video Generation (NeurIPS 2025 · similarity 0.84)
- H-MoRe: Learning Human-centric Motion Representation for Action Analysis (CVPR 2025 · similarity 0.83)
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.