Timm models default_cfg

vision
1

Hi there. I am currently using the beit model from timm as backbone. Here in the default_cfg it shows that the size should be 224x224x3, interpolation should be BICUBIC, mean and std should be 0.5. I am attaching a figure as well as shown below.

image
image2422×622 38 KB

So, my doubt is → should we write transforms like this (shown in the code below) to handle the default_config as shown in the image above or will the transforms like BICUBIC and normalize be handled automatically by the loaded timm model?

scale_transform = transforms.Resize(size=(224, 224), interpolation=transforms.InterpolationMode.BICUBIC)
channel_transform = transforms.ToTensor()
normalize = transforms.Normalize(mean=[0.5, 0.5, 0.5],
                                     std=[0.5, 0.5, 0.5])
composed_transform = transforms.Compose([scale_transform, channel_transform, normalize])

I am not able to find relevant results on the web. Please help.

Thanks,
Pradnesh