Research map / Detection and segmentation / CNN features
SAPNet++: Evolving Point-Prompted Instance Segmentation With Semantic and Spatial Awareness
Zhipeng Wei, Xumeng Han, Xuehui Yu, Xue Yang, Guorong Li, 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 point distance guidance (ablation) | SAPNet · COCO val2017 | mAP: 26.8 -> 27.5 | Improved |
| add selection refinement (ablation) | SAPNet · COCO val2017 | mAP: 27.5 -> 27.7 | Improved |
| add positive and negative proposals generator (ablation) | SAPNet · COCO val2017 | mAP: 27.7 -> 29.7 | Improved |
| add box mining strategy (ablation) | SAPNet · COCO val2017 | mAP: 29.7 -> 30.8 | Improved |
| add multi-mask proposal supervision (ablation) | SAPNet · COCO val2017 | mAP: 30.8 -> 31.2 | Improved |
| add negative proposals generator (ablation) | SAPNet · COCO val2017 | AP: 29.3 -> 30.7 | Improved |
| add positive proposals generator (ablation) | SAPNet · COCO val2017 | AP: 29.3 -> 29.8 | Improved |
| use both positive and negative generators (ablation) | SAPNet · COCO val2017 | AP: 29.3 -> 31.2 | Improved |
| set box mining thresholds Tmin1=0.6 and Tmin2=0.3 (ablation) | SAPNet · COCO val2017 | AP: 30.7 -> 31.2 | Improved |
| vary negative proposal IoU thresholds Tneg1 and Tneg2 (ablation) | SAPNet · COCO val2017 | AP: 30.8 -> 31.2 | No clear effect |
| upgrade SAPNet to SAPNet++ (self-distillation and affinity refinement) | SAPNet ResNet-50 1x · COCO val2017 | AP: 30.7 -> 32.4 | Improved |
| upgrade SAPNet to SAPNet++ with ResNet-101, 3x | SAPNet · COCO val2017 | AP: 34.6 -> 35.7 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- WISH — WISH: Weakly Supervised Instance Segmentation using Heterogeneous Labels (CVPR 2025) · 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
- Semantic-aware SAM for Point-Prompted Instance Segmentation (CVPR 2024 · similarity 0.92)
- P2Seg: Pointly-supervised Segmentation via Mutual Distillation (ICLR 2024 · similarity 0.90)
- PointINS: Point-Based Instance Segmentation (TPAMI 2021 · similarity 0.89)
- Bridge the Points: Graph-based Few-shot Segment Anything Semantically (NeurIPS 2024 · similarity 0.88)
- Extreme Point Supervised Instance Segmentation (CVPR 2024 · similarity 0.87)
- Curriculum Point Prompting for Weakly-Supervised Referring Image Segmentation (CVPR 2024 · similarity 0.87)
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.
