Research map / Detection and segmentation / Weakly supervised detection and segmentation
P2Object: Single Point Supervised Object Detection and Instance Segmentation
Pengfei Chen, Xuehui Yu, Xumeng Han, Kuiran Wang, Guorong Li, Lingxi Xie, Zhenjun Han, Jianbin Jiao
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 PBR stage losses (L_mil2 and L_neg) to CBP (ablation) | P2BNet · COCO-17 val | AP: 13.7 -> 21.7 | Improved |
| add negative loss L_neg on top of L_mil2 (ablation) | P2BNet · COCO-17 val | AP: 12.7 -> 21.7 | Improved |
| add L_mil2 in PBR without negative loss (ablation) | P2BNet · COCO-17 val | AP: 13.7 -> 12.7 | Hurt |
| use MIL loss L_mil1 instead of treating all bag proposals as positive (ablation) | P2BNet CBP · COCO-17 val | AP: 2.9 -> 13.7 | Improved |
| optimize PBR jointly with CBP (ablation) | P2BNet · COCO-17 train | mIoU_box: 50.2 -> 52.5 | Improved |
| replace basic sampling with SPSD sampling in the last PBR stage (P2BNet++) (ablation) | P2BNet-FR · COCO-17 val | AP: 22.1 -> 23.1 | Improved |
| add P2MNet (object-aware pixel perception with BSP) on P2BNet++ (ablation) | P2BNet++-FR · COCO-17 val | AP: 23.1 -> 25.9 | Improved |
| add P2MNet pixel perception for segmentation vs BoxInst on P2BNet++ pseudo boxes (ablation) | P2BNet++-BoxInst · COCO-17 val (segmentation) | AP: 21.0 -> 23.1 | Improved |
| retrain with Sparse R-CNN instead of Faster R-CNN | P2MNet · COCO-17 val | AP: 25.9 -> 29.6 | Improved |
| use COCO-pretrained detector instead of ImageNet | P2MNet-FR · Cityscapes val | AP: 25.0 -> 27.4 | Improved |
| upgrade P2BNet to P2BNet++ then P2MNet (ablation) | P2Object detector · VOC 2007 test | AP50: 60.4 -> 66.2 -> 68.2 | Improved |
| use P2MNet-MR instead of SAM-generated pseudo masks | Mask R-CNN · COCO-17 val (segmentation) | AP: 20.9 -> 23.1 | Improved |
Closest papers by meaning
- Point-to-Box Network for Accurate Object Detection via Single Point Supervision (ECCV 2022 · similarity 0.94)
- Pointly-Supervised Instance Segmentation (CVPR 2022 · similarity 0.89)
- Pseudo Mask Augmented Object Detection (CVPR 2018 · similarity 0.89)
- PointOBB-v2: Towards Simpler, Faster, and Stronger Single Point Supervised Oriented Object Detection (ICLR 2025 · similarity 0.89)
- Proposal-Free Network for Instance-Level Object Segmentation (TPAMI 2017 · similarity 0.89)
- Point2RBox: Combine Knowledge from Synthetic Visual Patterns for End-to-end Oriented Object Detection with Single Point Supervision (CVPR 2024 · similarity 0.89)
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.
