Research map / Detection and segmentation / CNN features
Pixel Consensus Voting for Panoptic Segmentation
Haochen Wang, Ruotian Luo, Michael Maire, Greg Shakhnarovich
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 |
|---|---|---|---|
| use segment loss normalization with λ=0.5 rather than pixel-averaged loss (ablation) | PCV · COCO val | PQ: 32.9 -> 37.5 | Improved |
| use segment loss normalization with λ=0.5 rather than pixel-averaged loss (ablation) | PCV · COCO val | PQth: 33.0 -> 40.0 | Improved |
| increase loss normalization strength from λ=0.5 to full area normalization, λ=1.0 (ablation) | PCV · COCO val | PQst: 33.7 -> 28.6 | Hurt |
| replace the default 233-cell voting grid with the simple grid (ablation) | PCV · COCO val | PQ: 37.5 -> 33.3 | Hurt |
| replace the default 233-cell voting grid with the simple grid (ablation) | PCV · COCO val oracle | PQ: 90.1 -> 79.2 | Hurt |
| replace the default voting grid with evenly spaced uniform bins (ablation) | PCV · COCO val oracle | PQ: 90.1 -> 67.1 | Hurt |
| replace the default voting grid with evenly spaced uniform bins (ablation) | PCV · Cityscapes val oracle | PQ: 88.6 -> 66.1 | Hurt |
| use PCV instead of the strongest listed baseline, UPSNet | panoptic segmentation · COCO val | PQ: 42.5 -> 37.5 | Hurt |
| use PCV instead of the strongest listed baseline, UPSNet | panoptic segmentation · Cityscapes val | PQ: 59.3 -> 54.2 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 3 on most shared results.
- SSAP — SSAP: Single-Shot Instance Segmentation With Affinity Pyramid (ICCV 2019) · better on 2 of 2 shared results
- UPSNet — UPSNet: A Unified Panoptic Segmentation Network (CVPR 2019) · better on 0 of 1 shared result
- AdaptIS — AdaptIS: Adaptive Instance Selection Network (ICCV 2019) · better on 0 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
- AdaptIS: Adaptive Instance Selection Network (ICCV 2019 · similarity 0.87)
- Unifying Training and Inference for Panoptic Segmentation (CVPR 2020 · similarity 0.87)
- A Generalist Framework for Panoptic Segmentation of Images and Videos (ICCV 2023 · similarity 0.87)
- Weakly- and Semi-Supervised Panoptic Segmentation (ECCV 2018 · similarity 0.87)
- Real-Time Panoptic Segmentation From Dense Detections (CVPR 2020 · similarity 0.87)
- Pixelwise Instance Segmentation With a Dynamically Instantiated Network (CVPR 2017 · 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.
