Research map / Detection and segmentation / Weakly supervised detection and segmentation
UFO²: A Unified Framework towards Omni-supervised Object Detection
Zhongzheng Ren, Zhiding Yu, Xiaodong Yang, Ming-Yu Liu, Alexander G. Schwing, Jan Kautz
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 teacher loss L_T1 (ablation) | UFO2 (boxes) · COCO-val, VGG-16 | AP: 22.6 -> 24.8 | Improved |
| add teacher loss L_T2 (ablation) | UFO2 (boxes) · COCO-val, VGG-16 | AP: 22.6 -> 25.1 | Improved |
| add L_T1 and L_T2 (ablation) | UFO2 (boxes) · COCO-val, VGG-16 | AP: 22.6 -> 25.5 | Improved |
| add tag teacher loss to L_T1 and L_T2 (ablation) | UFO2 (boxes) · COCO-val, VGG-16 | AP: 25.5 -> 25.7 | Improved |
| add proposal refinement for points (Ref_P) (ablation) | UFO2 (points) · COCO-val, VGG-16 | AP: 10.8 -> 11.1 | Improved |
| add proposal refinement for scribbles (Ref_S) (ablation) | UFO2 (scribbles) · COCO-val, VGG-16 | AP: 10.8 -> 11.6 | Improved |
| add localization constraint for points (Con_P) (ablation) | UFO2 (points) · COCO-val, VGG-16 | AP: 11.1 -> 12.4 | Improved |
| add localization constraint for scribbles (Con_S) (ablation) | UFO2 (scribbles) · COCO-val, VGG-16 | AP: 11.6 -> 13.7 | Improved |
| add unlabeled data to fine-tune weakly supervised model | UFO2 ResNet-50 (tags) · COCO minival | AP: 9.8 -> 10.5 | Improved |
| add scribbles to fine-tune box-supervised model | UFO2 ResNet-50 (boxes) · COCO minival | AP: 29.1 -> 30.9 | Improved |
| add tags to fine-tune box-supervised model | UFO2 ResNet-50 (boxes) · COCO minival | AP: 29.1 -> 29.4 | Improved |
| train with 80% boxes budget instead of mostly tags | UFO2 · COCO budget policy | AP: 3.0 ± 0.57 -> 14.11 ± 1.01 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- C-MIDN — C-MIDN: Coupled Multiple Instance Detection Network With Segmentation Guidance for Weakly Supervised Object Detection (ICCV 2019) · better on 1 of 1 shared result
- Faster-RCNN — Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks (TPAMI 2016) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Many-shot from Low-shot: Learning to Annotate using Mixed Supervision for Object Detection (ECCV 2020 · similarity 0.88)
- Co-Student: Collaborating Strong and Weak Students for Sparsely Annotated Object Detection (ECCV 2024 · similarity 0.88)
- Omni-DETR: Omni-Supervised Object Detection With Transformers (CVPR 2022 · similarity 0.88)
- AnnofreeOD: Detecting All Classes at Low Frame Rates Without Human Annotations (ICCV 2025 · similarity 0.87)
- Partial Weakly-Supervised Oriented Object Detection (CVPR 2026 · similarity 0.87)
- Object Detection Benchmarks are Incomplete: The Role of Label Errors and Annotation Uncertainty (NeurIPS 2026 · 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.
