Research map / Video / Action recognition
Social Scene Understanding: End-To-End Multi-Person Action Localization and Collective Activity Recognition
Timur Bagautdinov, Alexandre Alahi, Francois Fleuret, Pascal Fua, Silvio Savarese
Check 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 temporal matching RNN to the single-frame model (ablation) | OURS-single on Volleyball · Volleyball | collective activity accuracy with MRF detections: 83.3 -> 87.1 | Improved |
| Add temporal matching RNN to the single-frame model (ablation) | OURS-single on Volleyball · Volleyball | individual action accuracy with MRF detections: 77.8 -> 77.9 | No clear effect |
| Use ground-truth boxes rather than MRF detections (ablation) | OURS-temporal on Volleyball · Volleyball | individual action accuracy: 77.9 -> 82.4 | Improved |
| Use embedding-space nearest-neighbor matching instead of box-coordinate matching (ablation) | OURS-temporal with MRF detections · Volleyball | collective activity accuracy: 82.0 -> 87.1 | Improved |
| Use embedding-space nearest-neighbor matching instead of box-coordinate matching (ablation) | OURS-temporal with MRF detections · Volleyball | individual action accuracy: 68.6 -> 77.9 | Improved |
| Use soft embedding matching instead of embedding nearest-neighbor matching (ablation) | OURS-temporal with MRF detections · Volleyball | collective activity accuracy: 87.1 -> 86.2 | Hurt |
| Use soft embedding matching instead of embedding nearest-neighbor matching (ablation) | OURS-temporal with ground-truth detections · Volleyball | collective activity accuracy: 90.0 -> 90.6 | Improved |
| Replace NMS with MRF detection refinement (ablation) | OURS-temporal with boxes matching · Volleyball | collective activity accuracy: 77.0 -> 82.0 | Improved |
| Replace NMS with MRF detection refinement (ablation) | OURS-temporal with embedding matching · Volleyball | collective activity accuracy: 85.2 -> 87.1 | Improved |
| Replace NMS with MRF detection refinement (ablation) | OURS-temporal with soft embedding matching · Volleyball | individual action accuracy: 75.7 -> 77.4 | Improved |
| Use a shared multi-task representation rather than a person-only classifier | OURS-single with ground-truth boxes versus Inception-person · Volleyball | individual action accuracy: 78.1 -> 81.1 | Improved |
Closest papers by meaning
- Joint Learning of Social Groups, Individuals Action and Sub-group Activities in Videos (ECCV 2020 · similarity 0.91)
- JRDB-Act: A Large-Scale Dataset for Spatio-Temporal Action, Social Group and Activity Detection (CVPR 2022 · similarity 0.89)
- SocialGesture: Delving into Multi-person Gesture Understanding (CVPR 2025 · similarity 0.87)
- Learning to Evolve Scenes: Reasoning about Human Activities with Scene Graphs (NeurIPS 2026 · similarity 0.86)
- Social-Transmotion: Promptable Human Trajectory Prediction (ICLR 2024 · similarity 0.86)
- MTGS: A Novel Framework for Multi-Person Temporal Gaze Following and Social Gaze Prediction (NeurIPS 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.
