how to assure the identical transform on both image and labels

[ppjerry]

Hi, Thank for this great project. I have a question about the data augmentation. In Image_Dataset_folder class, the transforms of the image and label are separate. How to assure the same random transform of images could be used in the related labels?

[bigmb]

Hello,

The way they are saved and called in the project is my way to make sure that same label and image are called.
I have taken a '0000' formar to make sure that happens after transformation.

[ppjerry]

'0000', what is that?

[bigmb]

Sorry I forgot to add that in the code on how i save the files.
Instead of directly just increasing the value of 'i'. I saved it in a proper format of '0001'-'0002' , so the order of the images didnt change. I made that mistake initially and I had to save it in this format to avoid all this mixing of labels and images.
If you want that code, let me know, I will have to dig up some old files for that.

[ppjerry]

Hi, In the following code, the self.tx worked on the image and the self.lx worked on the label. There is a randomRotation transformation in self.tx but there is no same randomrotation in the self.lx. Does it mean that the image will have the random rotation transformation but its label will not be rotated?

    if self.transformI:
        self.tx = self.transformI
    else:
        self.tx = torchvision.transforms.Compose([
          #  torchvision.transforms.Resize((128,128)),
            torchvision.transforms.CenterCrop(96),
            torchvision.transforms.RandomRotation((-10,10)),
           # torchvision.transforms.RandomHorizontalFlip(),
            torchvision.transforms.ColorJitter(brightness=0.4, contrast=0.4, saturation=0.4),
            torchvision.transforms.ToTensor(),
            torchvision.transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])
        ])

    if self.transformM:
        self.lx = self.transformM
    else:
        self.lx = torchvision.transforms.Compose([
          #  torchvision.transforms.Resize((128,128)),
            torchvision.transforms.CenterCrop(96),
            torchvision.transforms.Grayscale(),
            torchvision.transforms.ToTensor(),
            #torchvision.transforms.Lambda(lambda x: torch.cat([x, 1 - x], dim=0))
        ])

def __len__(self):

    return len(self.images)

def __getitem__(self, i):
    i1 = Image.open(self.images_dir + self.images[i])
    l1 = Image.open(self.labels_dir + self.labels[i])

    return self.tx(i1), self.lx(l1)

[bigmb]

I just checked that out. And yes that's a mistake from my end.

I will correct it in some days, but if you do, send me a pull request.
Check out this issue for the solution:[https://github.com/pytorch/vision/issues/9]

[ppjerry]

I just had a pull request try

[bigmb]

Merged.