List your company

Semantic Segmentation on Cityscapes (val)

84.2mIoUSegMAN-L
79.72880.88982.0583.211Nov 27, 2018Jan 2, 2020Feb 7, 2021Mar 15, 2022Apr 21, 2023May 26, 2024Jul 2, 2025

Evaluation results

MethodLinks
SegMAN-LParams(M)=92.4, GFLOP=796.0, Resolution=2048x10242024.1284.2——————————————————————————————
SegFormerEncoder=MiT-B5, Params=84.7M, Flops=1447.6G, FPS=2.5, Category=Non Real-Time2021.0584——————————————————————————————
SeMask-L Mask2FormerBackbone=SeMask Swin-L†2021.1283.97———84.98——————————————————————————
SegFormerEncoder=MiT-B4, Params=64.1M, Flops=1240.6G, FPS=3.0, Category=Non Real-Time2021.0583.8——————————————————————————————
SegMAN-BParams(M)=51.8, GFLOP=479.0, Resolution=2048x10242024.1283.8——————————————————————————————
Mask2FormerBackbone=Swin-L2021.1283.3———84.3——————————————————————————
SegMAN-SParams(M)=29.4, GFLOP=218.4, Resolution=2048x10242024.1283.2——————————————————————————————
SegNeXt-LParams(M)=48.9, GFLOP=577.5, Resolution=2048x10242024.1283.2——————————————————————————————
HRViT-b3Backbone=HRViT-b3, #Param. (M)=28.6, GFLOPs=66.8, Decoder Head=SegFormer Head2021.1183.16——————————————————————————————
Panoptic-DeepLabBackbone=Xception-71, Extra Data=Mapillary Vistas, MS=true2019.1183.1——————————————————————————————
SpineNet-SegBackbone=SpineNet-S143+, Multi-scale Test=false2021.0383.04——————————————————————————————
HRViT-b2Backbone=HRViT-b2, #Param. (M)=20.8, GFLOPs=27.4, Decoder Head=SegFormer Head2021.1182.81——————————————————————————————
SenFormerbackbone=Swin-L2021.1182.8———84——————————————————————————
VWFormer-B5Params(M)=84.6, GFLOP=1139.6, Resolution=2048x10242024.1282.8——————————————————————————————
ResNeStBackbone=ResNeSt-200, Multi-scale Test=true2021.0382.7——————————————————————————————
Panoptic-DeepLabBackbone=Xception-71, Extra Data=Mapillary Vistas, Flip=true2019.1182.6——————————————————————————————
SegNeXt-BParams(M)=27.6, GFLOP=279.1, Resolution=2048x10242024.1282.6——————————————————————————————
CSWin-SBackbone=CSWin-S, #Param. (M)=37.3, GFLOPs=78.1, Decoder Head=SegFormer Head2021.1182.58——————————————————————————————
Panoptic-DeepLabBackbone=Xception-71, Extra Data=Mapillary Vistas2019.1182.5——————————————————————————————
SegFormerbackbone=MiT-B52021.1182.4———84——————————————————————————
VWFormer-B3Params(M)=47.3, GFLOP=637.1, Resolution=2048x10242024.1282.4——————————————————————————————
Segformer-B5Params(M)=84.7, GFLOP=1149.8, Resolution=2048x10242024.1282.4——————————————————————————————
MiT-B4Backbone=MiT-B4, #Param. (M)=64.1, GFLOPs=95.7, Decoder Head=SegFormer Head2021.1182.3——————————————————————————————
SETREncoder=ViT-Large, Params=318.3M, FPS=0.5, Category=Non Real-Time2021.0582.2——————————————————————————————
RPCNetBackbone=ResNet101, test_strategy=multi-scale test (MS) + Flip2020.0482.1——————————————————————————————
HANetBackbone=ResNet-1012020.0382.05——————————————————————————————
RPCNetBackbone=ResNet101, test_strategy=single scale test (SS) + Flip2020.0481.8——————————————————————————————
MiT-B3Backbone=MiT-B3, #Param. (M)=47.3, GFLOPs=79.0, Decoder Head=SegFormer Head2021.1181.7——————————————————————————————
VWFormer-B2Params(M)=27.4, GFLOP=415.1, Resolution=2048x10242024.1281.7——————————————————————————————
Segformer-B3Params(M)=47.3, GFLOP=962.9, Resolution=2048x10242024.1281.7——————————————————————————————
HRViT-b1Backbone=HRViT-b1, #Param. (M)=8.1, GFLOPs=14.1, Decoder Head=SegFormer Head2021.1181.63——————————————————————————————
HRNetV2 + OCRBackbone=HRNetV2-W48, #param.=70.3M, GFLOPS=1206.32019.0881.6——————————————————————————————
HRNetV2 + OCRBackbone=HRNetV2-W48, Multi-scale Test=true2021.0381.6——————————————————————————————
EDAFormer-BParams(M)=29.4, GFLOP=605.9, Resolution=2048x10242024.1281.6——————————————————————————————
CSWin-TBackbone=CSWin-T, #Param. (M)=22.4, GFLOPs=28.3, Decoder Head=SegFormer Head2021.1181.56——————————————————————————————
Panoptic-DeepLabBackbone=Xception-71, MS=true2019.1181.5——————————————————————————————
DANetBackbone=ResNet-101, Year=20192020.0381.5——————————————————————————————
DANetBackbone=ResNet1012020.0481.5——————————————————————————————
Panoptic-DeepLabMulti-scale Test=false2021.0381.5——————————————————————————————
SegFormer-B3Backbone=MiT-B3, Resolution=768x7682021.1281.5——————————————————————————————
FeedFormer-B2Params(M)=29.1, GFLOP=522.7, Resolution=2048x10242024.1281.5——————————————————————————————
ACFNetBackbone=ResNet-101, Year=20192020.0381.46——————————————————————————————
Baseline + VRec with JP + Label Relaxationsplit=val, Video Reconstruction=true, Joint Propagation=true, Label Relaxation=true2018.1281.4——————————98.587.493.564.266.169.374.281.592.964.695.683.566.595.787.791.985.770.178.8—
SynBoostAux data=true, Image Resynthesis=true2021.1281.4——————————————————————————————
Image Resyn.Aux data=false, Image Resynthesis=true2021.1281.4——————————————————————————————
CCNetBackbone=ResNet-101, Year=20192020.0381.3——————————————————————————————
HRNetV2Backbone=HRNetV2-W48, #param.=65.9M, GFLOPS=696.22019.0881.1——————————————————————————————
HRNetV2Backbone=HRNetV2-W48, Multi-scale Test=false2021.0381.1——————————————————————————————
OCRNetEncoder=HRNet-W48, Params=70.5M, Flops=1296.8G, FPS=4.2, Category=Non Real-Time2021.0581.1——————————————————————————————
Axial-DeepLabEncoder=AxialResNet-XL, Flops=2446.8G, Category=Non Real-Time2021.0581.1——————————————————————————————
LogicDiagBackbone=ResNet-101, Labeled Ratio=1/2, Number of Labeled Images=1488, Params=59.5M2025.0681——————————————————————————————
FARCLUSSBackbone=ResNet-101, Labeled Ratio=1/2, Number of Labeled Images=1488, Params=59.5M2025.0681——————————————————————————————
Axial-DLBackbone=Axial-L2021.0681——————————————————————————————
MiT-B2Backbone=MiT-B2, #Param. (M)=27.5, GFLOPs=62.4, Decoder Head=SegFormer Head2021.1181——————————————————————————————
Segformer-B2Params(M)=27.5, GFLOP=717.1, Resolution=2048x10242024.1281——————————————————————————————
Kirillov et al.Backbone=ResNet-50, Pre-training=ImageNet, Testing Scale=Multi-scale2019.0180.9——————————————————————————————
Panoptic-DeepLabBackbone=Xception-71, Flip=true2019.1180.9——————————————————————————————
DeeplabV3+Encoder=ResNet-101, Params=62.7M, Flops=2032.3G, FPS=1.2, Category=Non Real-Time2021.0580.9——————————————————————————————
DeepLabV3+Backbone=ResNet-101, Resolution=1024x5122021.1280.9——————————————————————————————
DeepLabV3+backbone=R1012021.1180.9———82.03——————————————————————————
DDFPBackbone=ResNet-101, Labeled Ratio=1/2, Number of Labeled Images=1488, Params=59.5M2025.0680.8——————————————————————————————
GSCNN2019.0780.8——————————98.386.393.355.86470.875.983.19365.195.285.367.99680.891.283.369.680.4—
GSCNNBackbone=ResNet1012020.0480.8——————————————————————————————
GSCNNEncoder=WideResNet38, Category=Non Real-Time2021.0580.8——————————————————————————————
SeamlessBackbone=Xception-71, Extra Data=Mapillary Vistas2019.1180.7——————————————————————————————
Dynamic RoutingEncoder=Dynamic-L33-PSP, Flops=270.0G, Category=Non Real-Time2021.0580.7——————————————————————————————
OCRNetBackbone=HRNetV2p-W48, Resolution=1024x5122021.1280.7——————————————————————————————
DAWBackbone=ResNet-101, Labeled Ratio=1/2, Number of Labeled Images=1488, Params=59.5M2025.0680.6——————————————————————————————
RepVGGBackbone=RepVGG-B2, Multi-scale Test=false2021.0380.57——————————————————————————————
ESLBackbone=ResNet-101, Labeled Ratio=1/2, Number of Labeled Images=1488, Params=59.5M2025.0680.5——————————————————————————————
Baseline + VRec with JPsplit=val, Video Reconstruction=true, Joint Propagation=true2018.1280.5——————————9886.594.747.667.169.671.880.492.258.495.688.371.195.676.884.790.379.680.3—
Panoptic-DeepLabBackbone=Xception-712019.1180.5——————————————————————————————
Panoptic-DLBackbone=Xception-712021.0680.5——————————————————————————————
SegFormer-B2Backbone=MiT-B2, Resolution=768x7682021.1280.5——————————————————————————————
CCNetBackbone=ResNet-1012021.1280.5———81.3——————————————————————————
AugSegBackbone=ResNet-101, Labeled Ratio=1/2, Number of Labeled Images=1488, Params=59.5M2025.0680.4——————————————————————————————
CorrMatchBackbone=ResNet-101, Labeled Ratio=1/2, Number of Labeled Images=1488, Params=59.5M2025.0680.4——————————————————————————————
DANetBackbone=ResNet-101, Resolution=1024x5122021.1280.4——————————————————————————————
DNLBackbone=ResNet-101, Resolution=1024x5122021.1280.4——————————————————————————————
Auto-DeepLab-LF=48, Multi-Adds=695.03B, Params=44.42M, ImageNet pre-trained=false, inference_scale=single-scale2019.0180.33——————————————————————————————
AutoDeeplabBackbone=AutoDeeplab-L, Multi-scale Test=false2021.0380.33——————————————————————————————
BiSeNetBackbone=ResNet-101, Multi-scale evaluation=true2018.1180.3——————————————————————————————
MDEQ-XLBackbone=MDEQ, Multi-scale Test=false2021.0380.3——————————————————————————————
Auto-DeeplabEncoder=NAS-F48-ASPP, Flops=695.0G, Category=Non Real-Time2021.0580.3——————————————————————————————
Max softmaxAux data=false, Image Resynthesis=false2021.1280.3——————————————————————————————
Embed. Dens.Aux data=false, Image Resynthesis=false2021.1280.3——————————————————————————————
SegMAN-TParams(M)=6.4, GFLOP=52.5, Resolution=2048x10242024.1280.3——————————————————————————————
LogicDiagBackbone=ResNet-101, Labeled Ratio=1/4, Number of Labeled Images=744, Params=59.5M2025.0680.2——————————————————————————————
DeeplabArchitecture=DeepLabV3+, Backbone=ResNet101, Supervision=Full supervision, CRF postprocessing=No, Training=Single-stage2025.0780.2——————————————————————————————
HRNetV2Backbone=HRNetV2-W40, #param.=45.2M, GFLOPS=493.22019.0880.2——————————————————————————————
CCNetEncoder=ResNet-101, Params=68.9M, Flops=2224.8G, FPS=1.0, Category=Non Real-Time2021.0580.2——————————————————————————————
CGRSeg-T†Params(M)=35.7, GFLOP=199.7, Resolution=2048x10242024.1280.2——————————————————————————————
DeepLabV3+Backbone=ResNet-50, Resolution=1024x5122021.1280.1——————————————————————————————
Auto-DeepLab-MF=32, Multi-Adds=460.93B, Params=21.62M, ImageNet pre-trained=false, inference_scale=single-scale2019.0180.04——————————————————————————————
FARCLUSSBackbone=ResNet-101, Labeled Ratio=1/4, Number of Labeled Images=744, Params=59.5M2025.0680——————————————————————————————
ShelfNet34-lwBackbone=ResNet-34, Multi-scale evaluation=true2018.1180——————————————————————————————
DDFPBackbone=ResNet-101, Labeled Ratio=1/4, Number of Labeled Images=744, Params=59.5M2025.0679.9——————————————————————————————
ANLNetBackbone=ResNet-101, Year=20192020.0379.9——————————————————————————————
SenFormerbackbone=R1012021.1179.9———81.4——————————————————————————
TDNet-PSPNetFLOPs (G)=5412022.0379.9——————————————————————————————