Heedless Backbones

CSWin Family

Select an option
Results
Parameters (M)
Images / Second
Publication Date
Select an option
---------
Object Detection
Instance Segmentation
Classification
Semantic Segmentation
Panoptic Segmentation
Select an option
---------
Cityscapes (val)
Cityscapes (test)
ADE20K (val)
ADE20K (test)
Select an option
mIoUms
pAccms
mAccms
mIoUss
pAccss
mAccss
GFLOPs
Select an option
---------
UPerNet
Mask2Former
Panoptic FPN
Select an option
----------
512x2048
640x2560
Select an option
Results
Parameters (M)
Images / Second
GFLOPs
Publication Date
Select an option
---------
JFT-3B
ImageNet-1k
ImageNet-22k
JFT-300M
Select an option
----------
Supervised
FCMAE
MAE
CL
Select an option
Family
Pretrain Dataset
Pretrain Method
Semantic Segmentation Head
Semantic Segmentation Resolution
Semantic Segmentation Training Epochs
Select an option
----------
Family
Pretrain Method
Semantic Segmentation Head
Semantic Segmentation Resolution
Semantic Segmentation Training Epochs
modelparams (m)pretrainheadtrainGFLOPsmIoUms
CSWin-T23.0IN-1k : Sup. : 300UPerNetADE20K (train) : 128 : 512959.050.7
CSWin-S35.0IN-1k : Sup. : 300UPerNetADE20K (train) : 128 : 5121027.051.5
CSWin-B78.0IN-1k : Sup. : 300UPerNetADE20K (train) : 128 : 5121222.052.2
CSWin-B78.0IN-22k : Sup. : 90UPerNetADE20K (train) : 128 : 6401941.052.6
CSWin-L173.0IN-22k : Sup. : 90UPerNetADE20K (train) : 128 : 6402745.055.7
modelparams (m)pretrainfinetunegflopsIN-1k
CSWin-T23.0IN-1k : Sup. : 300— : — : —4.382.7/—
CSWin-T23.0IN-1k : Sup. : 300IN-1k : 30 : 38414.084.3/—
CSWin-S35.0IN-1k : Sup. : 300— : — : —6.983.6/—
CSWin-S35.0IN-1k : Sup. : 300IN-1k : 30 : 38422.085.0/—
CSWin-B78.0IN-1k : Sup. : 300— : — : —15.084.2/—
CSWin-B78.0IN-1k : Sup. : 300IN-1k : 30 : 38447.085.4/—
CSWin-B78.0IN-22k : Sup. : 90IN-1k : 30 : 22415.085.9/—
CSWin-B78.0IN-22k : Sup. : 90IN-1k : 30 : 38447.087.0/—
CSWin-L173.0IN-22k : Sup. : 90IN-1k : 30 : 22431.586.5/—
CSWin-L173.0IN-22k : Sup. : 90IN-1k : 30 : 38496.887.5/—

COCO (val)

modelpretrainheadtraingflopsmAPbAPb50APb75mAPbsmAPbmmAPbl
CSWin-TIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 12279.046.768.651.3
CSWin-TIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 36279.049.070.753.7
CSWin-TIN-1k : Sup. : 300Cascade Mask R-CNNCOCO (train) : 36757.052.571.557.1
CSWin-SIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 12342.047.970.152.6
CSWin-SIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 36342.050.071.354.7
CSWin-SIN-1k : Sup. : 300Cascade Mask R-CNNCOCO (train) : 36820.053.772.258.4
CSWin-BIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 12526.048.770.453.9
CSWin-BIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 36526.050.872.155.8
CSWin-BIN-1k : Sup. : 300Cascade Mask R-CNNCOCO (train) : 361004.053.972.658.5

COCO (val)

modelpretrainheadtraingflopsmAPmAPm50APm75mAPmsmAPmmmAPml
CSWin-TIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 12279.042.265.645.4
CSWin-TIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 36279.043.667.946.6
CSWin-TIN-1k : Sup. : 300Cascade Mask R-CNNCOCO (train) : 36757.045.368.848.9
CSWin-SIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 12342.043.267.146.2
CSWin-SIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 36342.044.568.447.7
CSWin-SIN-1k : Sup. : 300Cascade Mask R-CNNCOCO (train) : 36820.046.469.650.6
CSWin-BIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 12526.043.967.847.3
CSWin-BIN-1k : Sup. : 300Mask R-CNNCOCO (train) : 36526.044.969.148.3
CSWin-BIN-1k : Sup. : 300Cascade Mask R-CNNCOCO (train) : 361004.046.470.050.4

ADE20K (val)

modelpretrainheadtraingflopsmIoUmspAccmsmAccmsmIoUsspAccssmAccss
CSWin-TIN-1k : Sup. : 300UPerNetADE20K (train) : 128 : 512959.050.749.3
CSWin-TIN-1k : Sup. : 300Panoptic FPNADE20K (train) : 64 : 512202.048.2
CSWin-SIN-1k : Sup. : 300UPerNetADE20K (train) : 128 : 5121027.051.550.4
CSWin-SIN-1k : Sup. : 300Panoptic FPNADE20K (train) : 64 : 512271.049.2
CSWin-BIN-1k : Sup. : 300UPerNetADE20K (train) : 128 : 5121222.052.251.1
CSWin-BIN-1k : Sup. : 300Panoptic FPNADE20K (train) : 64 : 512464.049.9
CSWin-BIN-22k : Sup. : 90UPerNetADE20K (train) : 128 : 6401941.052.651.8
CSWin-LIN-22k : Sup. : 90UPerNetADE20K (train) : 128 : 6402745.055.754.0