This repo is a collection of the challenging panoptic segmentation, including papers, codes, and tutorials.
Summarize in one sentence : Panoptic Segmentation proposes to solve the semantic segmentation(Stuff) and instance segmentation(Thing) in a unified and general manner.
Generally, the datasets which contains both semantic and instance annotations can be used to solve the challenging panoptic task;
- Cityscapes
- Mapillary Vistas
- ADE20K
- COCO-Panoptic
- BDD100K (the instance annotations are temporaily not released)
- PQ(Panoptic Quality)
- SQ & RQ(Segmentation Quality and Recognition Quality)
- Mapillary Panoptic
- COCO Panoptic
- | PQ | SQ | RQ | PQ_{Th} | SQ_{Th} | RQ_{Th} | PQ_{St} | SQ_{St} | RQ_{St} | E2E |
---|---|---|---|---|---|---|---|---|---|---|
Megvii(Face++) | 0.532 | 0.830 | 0.632 | 0.621 | 0.852 | 0.726 | 0.398 | 0.797 | 0.489 | False |
Panoptic FPN | 0.409 | 0.483 | 0.297 | True |
- Cityscapes Panoptic
1.Face++ Detection Team on Panoptic Segmentation
1.Panoptic Segmentation
- Paper : https://arxiv.org/abs/1801.00868
2.Learning to Fuse Things and Stuff (CVPR2019)
- Paper : https://arxiv.org/abs/1812.01192
3.Panoptic Segmentation with a Joint Semantic and Instance Segmentation Network
- Paper : https://arxiv.org/abs/1809.02110
4.Attention-guided Unified Network for Panoptic Segmentation (CVPR2019)
- Paper : https://arxiv.org/abs/1812.03904
- Code : Will released
5.Weakly- and Semi-Supervised Panoptic Segmentation (ECCV2018)
- Paper : https://arxiv.org/abs/1808.03575
- Code : https://github.com/qizhuli/Weakly-Supervised-Panoptic-Segmentation
6.Interactive Full Image Segmentation
- Paper : https://arxiv.org/abs/1812.01888
7.Panoptic Feature Pyramid Networks (CVPR2019)
- Paper : https://arxiv.org/abs/1901.02446
8.UPSNet: A Unified Panoptic Segmentation Network (CVPR2019)
9.Single Network Panoptic Segmentation for Street Scene Understanding
- Paper : https://arxiv.org/abs/1902.02678
10.DeeperLab: Single-Shot Image Parser (CVPR2019)
- Paper : https://arxiv.org/abs/1902.05093
11 An End-to-End Network for Panoptic Segmentation (CVPR2019)
- Paper : https://arxiv.org/abs/1903.05027
- Code : Will released
12 PanopticFusion: Online Volumetric Semantic Mapping at the Level of Stuff and Things (CVPR2019)