Research map / Detection and segmentation / Open-set and few-shot detection
Open-World Instance Segmentation: Exploiting Pseudo Ground Truth From Learned Pairwise Affinity
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 pairwise-affinity pseudo-GT masks to annotated training masks | Mask R-CNN / GGN · COCO Person to non-Person | AR@100: 4.9 -> 20.9 | Helped |
| Add pairwise-affinity pseudo-GT masks to annotated training masks | Mask R-CNN / GGN · COCO VOC to non-VOC | AR@100: 19.9 -> 28.7 | Helped |
| Add instance-level end-to-end training on pseudo-GT masks (ablation) | PA+Grouping · COCO Person to non-Person | AR@100: 14.6 -> 20.9 | Helped |
| Use OWT+gPb+UCM rather than CC or GBH grouping (ablation) | Pairwise-affinity grouping · Affinity-map grouping | Recall@all: 14.4 -> 24.2 | Helped |
| Add oriented watershed transform to WT+UCM (ablation) | Pairwise-affinity grouping · Affinity-map grouping | Recall@all: 23.6 -> 23.8 | Helped |
| Add globalized contour (gPb) to OWT+UCM (ablation) | Pairwise-affinity grouping · Affinity-map grouping | Recall@all: 23.8 -> 24.2 | Helped |
| Aggregate eight affinity channels by minimum (ablation) | PA predictor · VOC/non-VOC | AR@100: 22.8/19.3 | Mixed |
| Use one-channel mean-affinity prediction instead of eight-channel prediction (ablation) | PA predictor · VOC/non-VOC | AR@100: 23.1/18.5 | Mixed |
| Use PA rather than RGB, depth, edge, or MoCo features (ablation) | Open-world representation · COCO Person to non-Person | AR@100: 4.9; 10.9; 10.5; 10.7 -> 14.1 | Helped |
| Use PA-generated UCM pseudo-masks rather than edge-generated UCM pseudo-masks (ablation) | Open-world representation · COCO Person to non-Person | AR@100: 7.9 -> 20.9 | Helped |
| Combine PA and OLN objectness to rank pseudo-masks (ablation) | GGN · ADE20K | AR: 18.3 -> 21.0 | Helped |
| Combine PA and OLN objectness to rank pseudo-masks (ablation) | GGN · UVO | AR: 42.6 -> 43.4 | Helped |
Closest papers by meaning
- Open3DIS: Open-Vocabulary 3D Instance Segmentation with 2D Mask Guidance (CVPR 2024 · similarity 0.89)
- The Devil is in the Object Boundary: Towards Annotation-free Instance Segmentation using Foundation Models (ICLR 2024 · similarity 0.88)
- MaskClustering: View Consensus based Mask Graph Clustering for Open-Vocabulary 3D Instance Segmentation (CVPR 2024 · similarity 0.88)
- v-CLR: View-Consistent Learning for Open-World Instance Segmentation (CVPR 2025 · similarity 0.88)
- OpenWorldSAM: Extending SAM2 for Universal Image Segmentation with Language Prompts (NeurIPS 2025 · similarity 0.88)
- Training-Free Dataset Pruning for Instance Segmentation (ICLR 2025 · 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.