Research map / Video / Action recognition
JRDB-Act: A Large-Scale Dataset for Spatio-Temporal Action, Social Group and Activity Detection
Mahsa Ehsanpour, Fatemeh Saleh, Silvio Savarese, Ian Reid, Hamid Rezatofighi
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 geometric features to the BCE grouping baseline (ablation) | Social group detection · JRDB-Act validation set, ground-truth boxes | overall group AP: 41.4 -> 52.2 | Improved |
| Replace heuristic group cardinality with learned MSE cardinality (ablation) | Social group detection · JRDB-Act validation set, ground-truth boxes | overall group AP: 52.2 -> 56.8 | Improved |
| Add eigenvalue-based loss (ablation) | Social group detection · JRDB-Act validation set, ground-truth boxes | overall group AP: 56.8 -> 59.2 | Improved |
| Partition action classes into separate balanced losses (ablation) | Individual action detection · JRDB-Act validation set, ground-truth boxes | Action mAP: 8.0 -> 9.0 | Improved |
| Weight action losses by inverse class frequency instead of partitioning (ablation) | Individual action detection · JRDB-Act validation set, ground-truth boxes | Action mAP: 8.0 -> 8.1 | No clear effect |
| Use the proposed framework instead of the prior framework | JRDB-Act test set with Faster-RCNN boxes · JRDB-Act test set | overall group AP: 21.1 -> 29.1 | Improved |
| Use the proposed framework instead of the prior framework | JRDB-Act test set with Faster-RCNN boxes · JRDB-Act test set | Action mAP: 4.4 -> 5.3 | Improved |
| Use the proposed framework instead of the prior framework | JRDB-Act test set with Faster-RCNN boxes · JRDB-Act test set | G-Act mAP2: 1.3 -> 3.4 | Improved |
| Use MMPAT instead of Faster-RCNN detection boxes | Proposed framework · JRDB-Act test set | overall group AP: 29.1 -> 31.5 | Improved |
| Use MMPAT instead of Faster-RCNN detection boxes | Proposed framework · JRDB-Act test set | Action mAP: 5.3 -> 5.4 | No clear effect |
| Include difficult labels in evaluation rather than evaluating easy labels only (ablation) | Proposed framework · JRDB-Act test set | overall group AP: 31.3 -> 22.9 | Hurt |
Closest papers by meaning
- JRDB-Social: A Multifaceted Robotic Dataset for Understanding of Context and Dynamics of Human Interactions Within Social Groups (CVPR 2024 · similarity 0.92)
- Joint Learning of Social Groups, Individuals Action and Sub-group Activities in Videos (ECCV 2020 · similarity 0.90)
- Social Scene Understanding: End-To-End Multi-Person Action Localization and Collective Activity Recognition (CVPR 2017 · similarity 0.89)
- JRDB-Pose: A Large-Scale Dataset for Multi-Person Pose Estimation and Tracking (CVPR 2023 · similarity 0.89)
- Learning from Synthetic Human Group Activities (CVPR 2024 · similarity 0.88)
- AVA: A Video Dataset of Spatio-Temporally Localized Atomic Visual Actions (CVPR 2018 · similarity 0.88)
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.
