TasksSotADatasetsPapersMethodsSubmitAbout
Papers With Code 2

A community resource for machine learning research: papers, code, benchmarks, and state-of-the-art results.

Explore

Notable BenchmarksAll SotADatasetsPapersMethods

Community

Submit ResultsAbout

Data sourced from the PWC Archive (CC-BY-SA 4.0). Built by the community, for the community.

SotA/Computer Vision/Panoptic Segmentation/Cityscapes val

Panoptic Segmentation on Cityscapes val

Metric: mIoU (higher is better)

LeaderboardDataset
Loading chart...

Results

Submit a result
#Model↕mIoU▼Extra DataPaperDate↕Code
1EfficientPS (Cityscapes-fine)90.3NoEfficientPS: Efficient Panoptic Segmentation2020-04-05Code
2ViT-P (OneFormer, InternImage-H)85.4NoThe Missing Point in Vision Transformers for Uni...2025-05-26Code
3Panoptic-DeepLab (SWideRNet [1, 1, 4.5], Mapillary Vistas, multi-scale)85.3YesScaling Wide Residual Networks for Panoptic Segm...2020-11-23-
4OneFormer (ConvNeXt-L, single-scale, 512x1024, Mapillary Vistas-pretrained)84.6YesOneFormer: One Transformer to Rule Universal Ima...2022-11-10Code
5Axial-DeepLab-XL (Mapillary Vistas, multi-scale)84.6YesAxial-DeepLab: Stand-Alone Axial-Attention for P...2020-03-17Code
6Panoptic-DeepLab (SWideRNet [1, 1, 4.5], Mapillary Vistas, single-scale)84.6YesScaling Wide Residual Networks for Panoptic Segm...2020-11-23-
7OneFormer (ConvNeXt-XL, single-scale)83.6NoOneFormer: One Transformer to Rule Universal Ima...2022-11-10Code
8kMaX-DeepLab (single-scale)83.5NokMaX-DeepLab: k-means Mask Transformer2022-07-08Code
9DiNAT-L (Mask2Former)83.4NoDilated Neighborhood Attention Transformer2022-09-29Code
10OneFormer (DiNAT-L, single-scale)83.1NoOneFormer: One Transformer to Rule Universal Ima...2022-11-10Code
11OneFormer (ConvNeXt-L, single-scale)83NoOneFormer: One Transformer to Rule Universal Ima...2022-11-10Code
12AFF-Base (single-scale, point-based Mask2Former)83NoAutoFocusFormer: Image Segmentation off the Grid2023-04-24Code
13OneFormer (Swin-L, single-scale)83NoOneFormer: One Transformer to Rule Universal Ima...2022-11-10Code
14Mask2Former (Swin-L)82.9NoMasked-attention Mask Transformer for Universal ...2021-12-02Code
15AFF-Small (single-scale, point-based Mask2Former)82.2NoAutoFocusFormer: Image Segmentation off the Grid2023-04-24Code
16EfficientPS82.1YesEfficientPS: Efficient Panoptic Segmentation2020-04-05Code
17Panoptic-DeepLab (X71)81.5YesPanoptic-DeepLab: A Simple, Strong, and Fast Bas...2019-11-22Code
18CMT-DeepLab (MaX-S, single-scale, IN-1K)81.4NoCMT-DeepLab: Clustering Mask Transformers for Pa...2022-06-17Code
19Dynamically Instantiated Network (ResNet-101)79.8NoWeakly- and Semi-Supervised Panoptic Segmentation2018-08-10Code
20COPS (ResNet-50)79.3NoCombinatorial Optimization for Panoptic Segmenta...2021-06-06Code
21AdaptIS (ResNeXt-101)79.2NoAdaptIS: Adaptive Instance Selection Network2019-09-17-
22UPSNet (ResNet-101, multiscale)79.2YesUPSNet: A Unified Panoptic Segmentation Network2019-01-12Code
23TASCNet (ResNet-50, multi-scale)78YesLearning to Fuse Things and Stuff2018-12-04-
24UPSNet (ResNet-101)77.8YesUPSNet: A Unified Panoptic Segmentation Network2019-01-12Code
25TASCNet (ResNet-50)77.8YesLearning to Fuse Things and Stuff2018-12-04-
26AdaptIS (ResNet-101)77.2NoAdaptIS: Adaptive Instance Selection Network2019-09-17-
27Panoptic FPN (ResNet-101)75.7NoPanoptic Feature Pyramid Networks2019-01-08Code
28AUNet (ResNet-101-FPN)75.6NoAttention-guided Unified Network for Panoptic Se...2018-12-10-
29AdaptIS (ResNet-50)75.3NoAdaptIS: Adaptive Instance Selection Network2019-09-17-
30UPSNet (ResNet-50)75.2NoUPSNet: A Unified Panoptic Segmentation Network2019-01-12Code