Research map / Video / Action recognition
Videos as Space-Time Region Graphs
Xiaolong Wang, Abhinav Gupta
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 naive average pooling of proposal features (ablation) | I3D · Charades (val) | mAP (%): 31.8 -> 32.1 | No clear effect |
| add spatial-temporal graph GCN (ablation) | I3D · Charades (val) | mAP (%): 31.8 -> 34.2 | Improved |
| add similarity graph GCN (ablation) | I3D · Charades (val) | mAP (%): 31.8 -> 35.0 | Improved |
| combine both graphs in a joint GCN (ablation) | I3D · Charades (val) | mAP (%): 34.2 -> 36.2 | Improved |
| change proposals per frame from 50 to 25 or 100 (ablation) | I3D + Joint GCN · Charades (val) | mAP (%): 36.2 -> 35.9 | No clear effect |
| add Joint GCN to Non-local I3D (ablation) | Non-local I3D · Charades (val) | mAP (%): 33.5 -> 37.5 | Improved |
| use ResNet-101 larger backbone with GCN | I3D + GCN · Charades | mAP (%): 35.5 -> 39.1 | Improved |
| add Joint GCN (ablation) | I3D · Something-Something (val) | top-1 accuracy (%): 41.6 -> 43.3 | Improved |
| add Joint GCN (ablation) | I3D · Something-Something (val) | top-5 accuracy (%): 72.2 -> 75.1 | Improved |
| use spatial-temporal graph only instead of joint (ablation) | I3D + GCN · Something-Something (val) | top-1 accuracy (%): 43.3 -> 42.8 | Hurt |
| add Joint GCN to Non-local I3D (ablation) | Non-local I3D · Something-Something (val) | top-1 accuracy (%): 44.4 -> 46.1 | Improved |
| add GCN overhead | I3D · Charades (val) | FLOPs: 153 -> 158 | No clear effect |
Closest papers by meaning
- Representation Learning on Visual-Symbolic Graphs for Video Understanding (ECCV 2020 · similarity 0.90)
- Recurrent Space-time Graph Neural Networks (NeurIPS 2019 · similarity 0.89)
- Unified Graph Structured Models for Video Understanding (ICCV 2021 · similarity 0.89)
- Towards Scene Graph Anticipation (ECCV 2024 · similarity 0.88)
- Learning to Evolve Scenes: Reasoning about Human Activities with Scene Graphs (NeurIPS 2026 · similarity 0.87)
- HIG: Hierarchical Interlacement Graph Approach to Scene Graph Generation in Video Understanding (CVPR 2024 · similarity 0.87)
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.
