Open HAHAHAZY opened 1 week ago
The transforms.ToTensor()
method already includes the normalization process. Are you asking why standardization is not included in the data augmentation? Since this study deals with medical images, we did not use the commonly calculated mean and standard deviation from the ImageNet dataset for standardizing the images.
Why is the normalization operation canceled during data augmentation? Does this have any impact on the segmentation task? I saw that the official code of mocov3 has a normalize operation