List your company
ResearchFormerly WizWand

Semantic Segmentation on Cityscapes (CS) 1024x512 (val)

82.6mIoUMask2Former
66.27270.51174.7578.989Jul 25, 2022
Updated 8mo ago

Evaluation results

MethodLinks
Mask2FormerBackbone=Swin-S, FLOPs (G)=97.1, #P (M)=68.82022.0782.6
Mask2FormerBackbone=Swin-T, FLOPs (G)=73.3, #P (M)=47.42022.0781.7
Trans4PASS (S)Backbone=Trans4PASS (S), FLOPs (G)=20, #P (M)=252022.0781.1
SegFormer-B2Backbone=SegFormer-B2, FLOPs (G)=25.3, #P (M)=24.72022.0781
DeepLabV3+Backbone=ResNet-101, FLOPs (G)=254, #P (M)=60.22022.0780.9
OCRNetBackbone=HRNetV2p-W48, FLOPs (G)=163, #P (M)=70.42022.0780.7
Trans4PASS+ (S)Backbone=Trans4PASS+ (S), FLOPs (G)=19.8, #P (M)=252022.0780.7
DANetBackbone=ResNet-101, FLOPs (G)=289, #P (M)=66.52022.0780.4
DNLBackbone=ResNet-101, FLOPs (G)=286, #P (M)=66.72022.0780.4
Mask2FormerBackbone=SegFormer-B2, FLOPs (G)=68.7, #P (M)=44.62022.0780.2
ResNeStBackbone=ResNeSt-101, FLOPs (G)=283, #P (M)=69.82022.0779.6
Trans4PASS+* (S)Backbone=Trans4PASS+ (S), FLOPs (G)=63.2, #P (M)=44.9, Head=Mask2Former, DMLPv2=true2022.0779.6
SETR-PUPBackbone=Transformer-L, FLOPs (G)=417, #P (M)=3102022.0779.3
Trans4PASS (T)Backbone=Trans4PASS (T), FLOPs (G)=12, #P (M)=13.92022.0779.1
MaskFormerBackbone=SegFormer-B2, FLOPs (G)=52.8, #P (M)=392022.0778.8
PSPNetBackbone=ResNet-50, FLOPs (G)=179, #P (M)=46.62022.0778.6
OCRNetBackbone=HRNetV2p-W18, FLOPs (G)=53.6, #P (M)=12.12022.0778.6
Trans4PASS+ (T)Backbone=Trans4PASS+ (T), FLOPs (G)=11.7, #P (M)=142022.0778.6
SegFormer-B1Backbone=SegFormer-B1, FLOPs (G)=15.5, #P (M)=13.72022.0778.5
SETR-NaiveBackbone=Transformer-L, FLOPs (G)=363, #P (M)=3062022.0777.9
Trans4PASS+* (T)Backbone=Trans4PASS+ (T), FLOPs (G)=55.2, #P (M)=33.9, Head=Mask2Former, DMLPv2=true2022.0777.8
Mask2FormerBackbone=SegFormer-B1, FLOPs (G)=58.9, #P (M)=33.62022.0777.6
SETR-MLABackbone=Transformer-L, FLOPs (G)=367, #P (M)=3112022.0777.2
Semantic-FPNBackbone=ResNet-101, FLOPs (G)=64.9, #P (M)=47.52022.0775.8
SwiftNetBackbone=ResNet-18, FLOPs (G)=9.3, #P (M)=8.52022.0775.4
ERFNetBackbone=ERFNet, FLOPs (G)=14.7, #P (M)=2.12022.0772.1
FANetBackbone=ResNet-34, FLOPs (G)=8.12022.0771.3
Fast-SCNNBackbone=Fast-SCNN, FLOPs (G)=0.9, #P (M)=1.42022.0769.1
MaskFormerBackbone=SegFormer-B1, FLOPs (G)=42.9, #P (M)=27.92022.0766.9