Research map / Detection and segmentation / CNN features
SSAP: Single-Shot Instance Segmentation With Affinity Pyramid
Naiyu Gao, Yanhu Shan, Yupei Wang, Xin Zhao, Yinan Yu, Ming Yang, Kaiqi Huang
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 cascaded graph partition initialized from 1/16 resolution (ablation) | SSAP graph partition · Cityscapes val | AP (%): 28.9 -> 31.5 | Improved |
| initialize cascade from 1/64 instead of 1/16 (ablation) | SSAP graph partition · Cityscapes val | AP (%): 31.5 -> 30.9 | Hurt |
| add joint learning with hierarchical affinity pyramid (ablation) | SSAP · Cityscapes val | AP (%): 30.0 -> 31.3 | Improved |
| add joint learning with affinity (r=5) to semantic segmentation (ablation) | SSAP · Cityscapes val | mIoU (%): 74.5 -> 75.0 | Improved |
| add joint learning to single-scale features without dilation (ablation) | SSAP · Cityscapes val | mIoU (%): 74.5 -> 74.2 | Hurt |
| add joint learning to single-scale features with dilation (ablation) | SSAP · Cityscapes val | AP (%): 30.6 -> 30.8 | Improved |
| enlarge affinity window from r=5 to r=9 (ablation) | SSAP · Cityscapes val | AP (%): 31.3 -> 30.0 | Hurt |
| add long-range affinities A2 to A5 to A1 (ablation) | SSAP · Cityscapes val | AP (%): 25.7 -> 31.5 | Improved |
| add balanced training data by dropping 80% all-1 pixels (ablation) | SSAP · Cityscapes val | AP (%): 29.1 -> 30.0 | Improved |
| add 3x affinity loss on object pixels (ablation) | SSAP · Cityscapes val | AP (%): 30.0 -> 31.3 | Improved |
| use kernel size 5 instead of 3 (ablation) | SSAP · Cityscapes val | AP (%): 31.3 -> 31.5 | Improved |
| add segmentation refinement (ablation) | SSAP ResNet-50 · Cityscapes val | AP (%): 28.7 -> 31.5 | Improved |
Later papers that compared with this method
7 later accepted papers report SSAP as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- Unsupervised Segmentation in Real-World Images via Spelke Object Inference (ECCV 2022 · better on 3 of 4 shared results)
- Split-GCN: Effective Interactive Annotation for Segmentation of Disconnected Instance (TPAMI 2022 · better on 0 of 1 shared result)
- Hierarchical Lovasz Embeddings for Proposal-Free Panoptic Segmentation (CVPR 2021 · better on 2 of 2 shared results)
- Pixel Consensus Voting for Panoptic Segmentation (CVPR 2020 · better on 2 of 2 shared results)
- Joint Semantic Instance Segmentation on Graphs with the Semantic Mutex Watershed (ECCV 2020 · better on 0 of 2 shared results)
- Unifying Training and Inference for Panoptic Segmentation (CVPR 2020 · better on 1 of 1 shared result)
- LevelSet R-CNN: A Deep Variational Method for Instance Segmentation (ECCV 2020 · better on 1 of 1 shared result)
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.
- PANet — PANet: Few-Shot Image Semantic Segmentation With Prototype Alignment (ICCV 2019) · better on 3 of 4 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
- Affinity Derivation and Graph Merge for Instance Segmentation (ECCV 2018 · similarity 0.91)
- Learning to Model Pixel-Embedded Affinity for Homogeneous Instance Segmentation (AAAI 2022 · similarity 0.90)
- Pixelwise Instance Segmentation With a Dynamically Instantiated Network (CVPR 2017 · similarity 0.90)
- Object Affinity Learning: Towards Annotation-Free Instance Segmentation (TPAMI 2023 · similarity 0.90)
- Semantic-aware SAM for Point-Prompted Instance Segmentation (CVPR 2024 · similarity 0.87)
- P2Seg: Pointly-supervised Segmentation via Mutual Distillation (ICLR 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.
