Research map / Detection and segmentation / Weakly supervised detection and segmentation
What’s the Point: Semantic Segmentation with Point Supervision
Amy Bearman, Olga Russakovsky, Vittorio Ferrari, Li Fei-Fei
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 objectness prior to image-level model (ablation) | Img (image-level) model · PASCAL VOC 2012 val | mIOU: 29.8 -> 32.2 | Improved |
| add one point per class supervision to image-level loss (ablation) | Img model · PASCAL VOC 2012 val | mIOU: 29.8 -> 35.1 | Improved |
| add objectness prior to point-supervised model (ablation) | 1Point Img model · PASCAL VOC 2012 val | mIOU: 35.1 -> 42.7 | Improved |
| add one point per class to Img + Obj model (ablation) | Img + Obj model · PASCAL VOC 2012 val | mIOU: 32.2 -> 42.7 | Improved |
| annotate all instances (AllPoints) instead of one point per class (ablation) | 1Point Img + Obj · PASCAL VOC 2012 val | mIOU: 42.7 -> 42.7 | No clear effect |
| weight all-instance points by rank (1/2^r) (ablation) | AllPoints Img + Obj · PASCAL VOC 2012 val | mIOU: 42.7 -> 43.4 | Improved |
| enlarge supervised area around each point to radius 2, 5, 25 pixels (ablation) | 1Point Img + Obj · PASCAL VOC 2012 val | mIOU: 43.0-43.1 | No clear effect |
| collect points from 3 annotators (ablation) | 1Point Img + Obj · PASCAL VOC 2012 val | mIOU: 42.7 -> 43.8 | Improved |
| use a different random annotator per image (ablation) | 1Point Img + Obj · PASCAL VOC 2012 val | mIOU: 42.8, 43.4, 43.8 vs 42.7 | No clear effect |
| replace human-clicked points with random points sampled on the object from ground truth (ablation) | 1Point Img + Obj · PASCAL VOC 2012 val | mIOU: 42.7 -> 46.1 | Improved |
| add 100 fully supervised images to point supervision (hybrid) | 1Point Img + Obj · PASCAL VOC 2012 val | mIOU: 42.7 -> 53.1 | Improved |
| replace points with one squiggle per class | Img + Obj · PASCAL VOC 2012 val | mIOU: 42.7 -> 49.1 | Improved |
Closest papers by meaning
- Single-Stage Semantic Segmentation From Image Labels (CVPR 2020 · similarity 0.90)
- Seed, Expand and Constrain: Three Principles for Weakly-Supervised Image Segmentation (ECCV 2016 · similarity 0.89)
- Region-Based Semantic Segmentation with End-to-End Training (ECCV 2016 · similarity 0.89)
- Curriculum Point Prompting for Weakly-Supervised Referring Image Segmentation (CVPR 2024 · similarity 0.88)
- Semantic-aware SAM for Point-Prompted Instance Segmentation (CVPR 2024 · similarity 0.88)
- Learning Hierarchical Image Segmentation For Recognition and By Recognition (ICLR 2024 · 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.
