Research map / Detection and segmentation / CNN features
Affinity Derivation and Graph Merge for Instance Segmentation
Yiding Liu, Siyu Yang, Bin Li, Wengang Zhou, Jizheng Xu, Houqiang Li, Yan Lu
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 pixel affinity refinement with semantic scores (ablation) | affinity + graph merge · Cityscapes val | AP: 18.9 -> 22.8 | Improved |
| add ROI resizing to height 513 (ablation) | affinity + graph merge · Cityscapes val | AP: 22.8 -> 28.7 | Improved |
| force local merge with window size 2 (ablation) | affinity + graph merge · Cityscapes val | AP: 28.7 -> 29.2 | Improved |
| force local merge with window size 4 (ablation) | affinity + graph merge · Cityscapes val | AP: 28.7 -> 27.5 | Hurt |
| add semantic class partition (ablation) | affinity + graph merge · Cityscapes val | AP: 29.2 -> 30.7 | Improved |
| finetune semantic branch with output stride 8 | affinity + graph merge · Cityscapes val | AP: 30.7 -> 31.2 | Improved |
| finetune instance branch with output stride 8 | affinity + graph merge · Cityscapes val | AP: 30.7 -> 31.2 | Improved |
| use output stride 8 for both branches | affinity + graph merge · Cityscapes val | AP: 30.7 -> 32.1 | Improved |
| add horizontal flip for semantic inference | affinity + graph merge · Cityscapes val | AP: 32.1 -> 32.8 | Improved |
| add horizontal flip for instance inference | affinity + graph merge · Cityscapes val | AP: 32.1 -> 32.6 | Improved |
| add both flips | affinity + graph merge · Cityscapes val | AP: 32.1 -> 33.5 | Improved |
| add semantic class refinement for bicycle and motorcycle | affinity + graph merge · Cityscapes val | AP: 33.5 -> 34.1 | 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.
- InstanceCut — InstanceCut: From Edges to Instances With MultiCut (CVPR 2017) · better on 2 of 2 shared results
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
- SSAP: Single-Shot Instance Segmentation With Affinity Pyramid (ICCV 2019 · similarity 0.91)
- Pixelwise Instance Segmentation With a Dynamically Instantiated Network (CVPR 2017 · similarity 0.89)
- Object Affinity Learning: Towards Annotation-Free Instance Segmentation (TPAMI 2023 · similarity 0.89)
- Learning to Model Pixel-Embedded Affinity for Homogeneous Instance Segmentation (AAAI 2022 · similarity 0.88)
- InstanceCut: From Edges to Instances With MultiCut (CVPR 2017 · similarity 0.87)
- SOLO: Segmenting Objects by Locations (ECCV 2020 · 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.
