Research map / Detection and segmentation / Weakly supervised detection and segmentation
Object Detection with a Unified Label Space from Multiple Datasets
Xiangyun Zhao, Samuel Schulter, Gaurav Sharma, Yi-Hsuan Tsai, Manmohan Chandraker, Ying Wu
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 pseudo labeling instead of UOD+Merge | unified detector · Ablation setup, MIX | AP50: 46.1 to 52.2 | Improved |
| add pseudo labeling instead of ignoring label space ambiguity (ablation) | unified detector · Ablation setup, V-on-C | AP50: 36.5 -> 62.2 | Improved |
| add partial-annotation sum loss instead of ignoring ambiguity (ablation) | unified detector · Ablation setup, MIX | AP50: 43.7 -> 44.1 | No clear effect |
| add minimum-entropy regularization to partial-loss sum (ablation) | unified detector · Ablation setup, MIX | AP50: 44.1 -> 43.9 | No clear effect |
| replace sum with max in partial-annotation loss (ablation) | unified detector · Ablation setup, MIX | AP50: 44.1 -> 44.6 | No clear effect |
| use hard-thresholding instead of soft-thresholding in the pseudo-label weighting (ablation) | pseudo labeling · Ablation setup, MIX | AP50: 51.6 -> 52.2 | Improved |
| add regression loss on pseudo ground truth (ablation) | pseudo labeling · Ablation setup, MIX | AP50: 52.0 -> 52.2 | Improved |
| use shared-backbone UOD+Merge instead of individual detectors (ablation) | UOD · Ablation setup, MIX | AP50: 46.9 -> 46.1 | No clear effect |
| threshold pseudo detections to confident ones only (ablation) | pseudo labeling of VOC detector on COCO train · COCO train pseudo labels | precision / recall: precision 0.29 -> 0.79, recall 0.7 -> 0.4 | Mixed |
| add pseudo labeling instead of UOD+Merge | unified detector · Setting A | AP50: 59.3 -> 61.1 | Improved |
| add pseudo labeling instead of UOD+Merge | unified detector · Setting B | AP50: 43.7 -> 48.5 | Improved |
| add pseudo labeling instead of partial-annotation loss | unified detector · Setting B | AP50: 41.2 -> 48.5 | Improved |
Closest papers by meaning
- Mind the Gap: Transferring Labels to Align Object Detection Datasets (CVPR 2026 · similarity 0.89)
- Transferring Labels to Solve Annotation Mismatches Across Object Detection Datasets (ICLR 2024 · similarity 0.89)
- Simple Multi-Dataset Detection (CVPR 2022 · similarity 0.89)
- Object Detection Benchmarks are Incomplete: The Role of Label Errors and Annotation Uncertainty (NeurIPS 2026 · similarity 0.88)
- Co-Student: Collaborating Strong and Weak Students for Sparsely Annotated Object Detection (ECCV 2024 · similarity 0.88)
- UniT: Unified Knowledge Transfer for Any-Shot Object Detection and Segmentation (CVPR 2021 · 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.
