Research map / Detection and segmentation / Open-set and few-shot detection
Relational Learning for Joint Head and Human Detection
Read the paperCheck 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 |
|---|---|---|---|
| Use a joint head-human framework with body proposals derived from head proposals instead of separate detectors | FPN head and human detection · CrowdHuman validation | Head MR^-2: 48.9% to 48.7% | Helped |
| Use a joint head-human framework with body proposals derived from head proposals instead of separate detectors | FPN head and human detection · CrowdHuman validation | Human MR^-2: 49.7% to 47.0% | Helped |
| Use the cascade branch in the separate human detector | FPN-Human · CrowdHuman validation | Human MR^-2: 49.7 -> 49.2 | Helped |
| Add the RDM to the joint detector to use head-body relations to remove false head detections (ablation) | JointDet · CrowdHuman validation | Head MR^-2: 48.7% to 48.3% | Helped |
| Add the RDM to the joint detector to recall bodies suppressed by NMS (ablation) | JointDet · CrowdHuman validation | Human MR^-2: 47.0% to 46.5% | Helped |
| Replace separate FPN detection with JointDet | Pedestrian detector · CityPersons validation Reasonable | MR^-2: 11.00% to 10.23% | Helped |
| Use the new re-annotated labels for training instead of refined annotations | FPN · Caltech-USA refined validation | MR^-2: 4.31% to 3.26% | Helped |
| Evaluate using new annotations rather than refined annotations | FPN trained with refined annotations · Caltech-USA validation | MR^-2: 4.31% to 16.52% | Hurt |
| Evaluate using new annotations rather than refined annotations | FPN trained with new annotations · Caltech-USA validation | MR^-2: 3.26% to 14.26% | Hurt |
Closest papers by meaning
- PairDETR : Joint Detection and Association of Human Bodies and Faces (CVPR 2024 · similarity 0.87)
- Revisiting the Sibling Head in Object Detector (CVPR 2020 · similarity 0.84)
- DeepCut: Joint Subset Partition and Labeling for Multi Person Pose Estimation (CVPR 2016 · similarity 0.83)
- Rethinking Classification and Localization for Object Detection (CVPR 2020 · similarity 0.83)
- SapiensID: Foundation for Human Recognition (CVPR 2025 · similarity 0.82)
- PBADet: A One-Stage Anchor-Free Approach for Part-Body Association (ICLR 2024 · similarity 0.82)
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.