It seems that the weight of VSCode-T and VSCode-S in model_zoo are exactly same

[lonelyqian]

args = parser.parse_args()
netT = ImageDepthNet(args)
netS = ImageDepthNet(args)
modelT_path = '/media/data/QFS/Models/proj-based/VScode-swin/RGB_VST_T.pth'
modelS_path = '/media/data/QFS/Models/proj-based/VScode-swin/RGB_VST_S.pth'
state_dictT = torch.load(modelT_path)
state_dictS = torch.load(modelS_path)
print(str(state_dictT) == str(state_dictS))

# True

The weitght of VSCode-T you provided is based on Swin-S not Swin-T

I wonder if you could please provide the pretrained Swin backbone?

[Sssssuperior]

I apologize for my error.
The VSCode-T version has been updated.

[typist2001]

请问Swin-T版本更新了吗？我百度网盘下载的Swin-T和Swin-S好像还是一样的