Semantic Segmentation on Cityscapes val

Metric: AP (higher is better)

LeaderboardDataset

Loading chart...

Results

Hide extra data

Sort:

#	Model↕	AP▼	Extra Data	Paper	Date↕	Code
1	ViT-P (OneFormer, InternImage-H)	50.6	No	The Missing Point in Vision Transformers for Uni...	2025-05-26	Code
2	OneFormer (ConvNeXt-L, single-scale, 512x1024, Mapillary Vistas-pretrained)	48.7	Yes	OneFormer: One Transformer to Rule Universal Ima...	2022-11-10	Code
3	Panoptic-DeepLab (SWideRNet [1, 1, 4.5], Mapillary Vistas, multi-scale)	46.8	Yes	Scaling Wide Residual Networks for Panoptic Segm...	2020-11-23	-
4	OneFormer (ConvNeXt-XL, single-scale)	46.7	No	OneFormer: One Transformer to Rule Universal Ima...	2022-11-10	Code
5	OneFormer (ConvNeXt-L, single-scale)	46.5	No	OneFormer: One Transformer to Rule Universal Ima...	2022-11-10	Code
6	AFF-Base (single-scale, point-based Mask2Former)	46.2	No	AutoFocusFormer: Image Segmentation off the Grid	2023-04-24	Code
7	OneFormer (DiNAT-L, single-scale)	45.6	No	OneFormer: One Transformer to Rule Universal Ima...	2022-11-10	Code
8	OneFormer (Swin-L, single-scale)	45.6	No	OneFormer: One Transformer to Rule Universal Ima...	2022-11-10	Code
9	DiNAT-L (Mask2Former)	44.5	No	Dilated Neighborhood Attention Transformer	2022-09-29	Code
10	Axial-DeepLab-XL (Mapillary Vistas, multi-scale)	44.2	Yes	Axial-DeepLab: Stand-Alone Axial-Attention for P...	2020-03-17	Code
11	AFF-Small (single-scale, point-based Mask2Former)	44.2	No	AutoFocusFormer: Image Segmentation off the Grid	2023-04-24	Code
12	kMaX-DeepLab (single-scale)	44	No	kMaX-DeepLab: k-means Mask Transformer	2022-07-08	Code
13	Mask2Former (Swin-L)	43.6	No	Masked-attention Mask Transformer for Universal ...	2021-12-02	Code
14	EfficientPS	43.5	Yes	EfficientPS: Efficient Panoptic Segmentation	2020-04-05	Code
15	Panoptic-DeepLab (SWideRNet [1, 1, 4.5], Mapillary Vistas, single-scale)	42.8	Yes	Scaling Wide Residual Networks for Panoptic Segm...	2020-11-23	-
16	EfficientPS (Cityscapes-fine)	39.1	No	EfficientPS: Efficient Panoptic Segmentation	2020-04-05	Code
17	UPSNet (ResNet-101, multiscale)	39	Yes	UPSNet: A Unified Panoptic Segmentation Network	2019-01-12	Code
18	TASCNet (ResNet-50, multi-scale)	39	Yes	Learning to Fuse Things and Stuff	2018-12-04	-
19	Panoptic-DeepLab (X71)	38.5	Yes	Panoptic-DeepLab: A Simple, Strong, and Fast Bas...	2019-11-22	Code
20	UPSNet (ResNet-101)	37.8	Yes	UPSNet: A Unified Panoptic Segmentation Network	2019-01-12	Code
21	TASCNet (ResNet-50)	37.6	Yes	Learning to Fuse Things and Stuff	2018-12-04	-
22	MRCNN + PSPNet (ResNet-101)	36.4	Yes	Panoptic Segmentation	2018-01-03	Code
23	AdaptIS (ResNeXt-101)	36.3	No	AdaptIS: Adaptive Instance Selection Network	2019-09-17	-
24	AUNet (ResNet-101-FPN)	34.4	No	Attention-guided Unified Network for Panoptic Se...	2018-12-10	-
25	COPS (ResNet-50)	34.1	No	Combinatorial Optimization for Panoptic Segmenta...	2021-06-06	Code
26	AdaptIS (ResNet-101)	33.9	No	AdaptIS: Adaptive Instance Selection Network	2019-09-17	-
27	UPSNet (ResNet-50)	33.3	No	UPSNet: A Unified Panoptic Segmentation Network	2019-01-12	Code
28	Panoptic FPN (ResNet-101)	33	No	Panoptic Feature Pyramid Networks	2019-01-08	Code
29	AdaptIS (ResNet-50)	32.3	No	AdaptIS: Adaptive Instance Selection Network	2019-09-17	-
30	Dynamically Instantiated Network (ResNet-101)	28.6	No	Weakly- and Semi-Supervised Panoptic Segmentation	2018-08-10	Code

#1ViT-P (OneFormer, InternImage-H)SOTA
50.6
AP· 2025-05-26
The Missing Point in Vision Transformers for Universal Image Segmentation Code
#2OneFormer (ConvNeXt-L, single-scale, 512x1024, Mapillary Vistas-pretrained)SOTA
48.7
AP· Extra Data· 2022-11-10
OneFormer: One Transformer to Rule Universal Image Segmentation Code
#3Panoptic-DeepLab (SWideRNet [1, 1, 4.5], Mapillary Vistas, multi-scale)SOTA
46.8
AP· Extra Data· 2020-11-23
Scaling Wide Residual Networks for Panoptic Segmentation
#4OneFormer (ConvNeXt-XL, single-scale)
46.7
AP· 2022-11-10
OneFormer: One Transformer to Rule Universal Image Segmentation Code
#5OneFormer (ConvNeXt-L, single-scale)
46.5
AP· 2022-11-10
OneFormer: One Transformer to Rule Universal Image Segmentation Code
#6AFF-Base (single-scale, point-based Mask2Former)
46.2
AP· 2023-04-24
AutoFocusFormer: Image Segmentation off the Grid Code
#7OneFormer (DiNAT-L, single-scale)
45.6
AP· 2022-11-10
OneFormer: One Transformer to Rule Universal Image Segmentation Code
#8OneFormer (Swin-L, single-scale)
45.6
AP· 2022-11-10
OneFormer: One Transformer to Rule Universal Image Segmentation Code
#9DiNAT-L (Mask2Former)
44.5
AP· 2022-09-29
Dilated Neighborhood Attention Transformer Code
#10Axial-DeepLab-XL (Mapillary Vistas, multi-scale)SOTA
44.2
AP· Extra Data· 2020-03-17
Axial-DeepLab: Stand-Alone Axial-Attention for Panoptic Segmentation Code
#11AFF-Small (single-scale, point-based Mask2Former)
44.2
AP· 2023-04-24
AutoFocusFormer: Image Segmentation off the Grid Code
#12kMaX-DeepLab (single-scale)
44
AP· 2022-07-08
kMaX-DeepLab: k-means Mask Transformer Code
#13Mask2Former (Swin-L)
43.6
AP· 2021-12-02
Masked-attention Mask Transformer for Universal Image Segmentation Code
#14EfficientPS
43.5
AP· Extra Data· 2020-04-05
EfficientPS: Efficient Panoptic Segmentation Code
#15Panoptic-DeepLab (SWideRNet [1, 1, 4.5], Mapillary Vistas, single-scale)
42.8
AP· Extra Data· 2020-11-23
Scaling Wide Residual Networks for Panoptic Segmentation
#16EfficientPS (Cityscapes-fine)
39.1
AP· 2020-04-05
EfficientPS: Efficient Panoptic Segmentation Code
#17UPSNet (ResNet-101, multiscale)
39
AP· Extra Data· 2019-01-12
UPSNet: A Unified Panoptic Segmentation Network Code
#18TASCNet (ResNet-50, multi-scale)SOTA
39
AP· Extra Data· 2018-12-04
Learning to Fuse Things and Stuff
#19Panoptic-DeepLab (X71)
38.5
AP· Extra Data· 2019-11-22
Panoptic-DeepLab: A Simple, Strong, and Fast Baseline for Bottom-Up Panoptic Segmentation Code
#20UPSNet (ResNet-101)
37.8
AP· Extra Data· 2019-01-12
UPSNet: A Unified Panoptic Segmentation Network Code
#21TASCNet (ResNet-50)
37.6
AP· Extra Data· 2018-12-04
Learning to Fuse Things and Stuff
#22MRCNN + PSPNet (ResNet-101)SOTA
36.4
AP· Extra Data· 2018-01-03
Panoptic Segmentation Code
#23AdaptIS (ResNeXt-101)
36.3
AP· 2019-09-17
AdaptIS: Adaptive Instance Selection Network
#24AUNet (ResNet-101-FPN)
34.4
AP· 2018-12-10
Attention-guided Unified Network for Panoptic Segmentation
#25COPS (ResNet-50)
34.1
AP· 2021-06-06
Combinatorial Optimization for Panoptic Segmentation: A Fully Differentiable Approach Code
#26AdaptIS (ResNet-101)
33.9
AP· 2019-09-17
AdaptIS: Adaptive Instance Selection Network
#27UPSNet (ResNet-50)
33.3
AP· 2019-01-12
UPSNet: A Unified Panoptic Segmentation Network Code
#28Panoptic FPN (ResNet-101)
33
AP· 2019-01-08
Panoptic Feature Pyramid Networks Code
#29AdaptIS (ResNet-50)
32.3
AP· 2019-09-17
AdaptIS: Adaptive Instance Selection Network
#30Dynamically Instantiated Network (ResNet-101)
28.6
AP· 2018-08-10
Weakly- and Semi-Supervised Panoptic Segmentation Code