Torchvision.transfors: How to perform identical transform on both image and target?


(KAI ZHAO) #1

How can I perform an identical transform on both image and target?

For example, in Semantic segmentation and Edge detection where the input image and target ground-truth are both 2D images, one must perform the same transform on both input image and target ground-truth.


(jdhao) #2

A simple search on Google will give you this


(KAI ZHAO) #3

Oh, my fault.
Thank you for the tips.


(KAI ZHAO) #4

Hi jdhao,

Sorry to bother you again.
I’ve checked related issues and prs, but I’m still confused about how to use the identical transform.

So is there any documentation about this part?

Thank you.


(jdhao) #5

PyTorch has an excellent tutorial on data loading. In that tutorial, the author shows how to do transform for both data and target. You can try to mimic the way in the tutorial.


(Issam H Laradji) #6

The question hasn’t been answered. The provided references don’t show the ideal practice to do identical transforms for both input image and segmentation label…


#7

Alternatively to the functions from the tutorial, you could use torchvision’s functional API.
Here is a small example for an image and the corresponding mask image:

class MyDataset(Dataset):
    def __init__(self, image_paths, target_paths, train=True):
        self.image_paths = image_paths
        self.target_paths = target_paths

    def transform(self, image, mask):
        # Resize
        resize = transforms.Resize(size=(520, 520))
        image = resize(image)
        mask = resize(mask)

        # Random crop
        i, j, h, w = transforms.RandomCrop.get_params(
            image, output_size=(512, 512))
        image = TF.crop(image, i, j, h, w)
        mask = TF.crop(mask, i, j, h, w)

        # Random horizontal flipping
        if random.random() > 0.5:
            image = TF.hflip(image)
            mask = TF.hflip(mask)

        # Random vertical flipping
        if random.random() > 0.5:
            image = TF.vflip(image)
            mask = TF.vflip(mask)

        # Transform to tensor
        image = TF.to_tensor(image)
        mask = TF.to_tensor(mask)
        return image, mask

    def __getitem__(self, index):
        image = Image.open(self.image_paths[index])
        mask = Image.open(self.target_paths[index])
        x, y = self.transform(image, mask)
        return x, y

    def __len__(self):
        return len(self.image_paths)

How to random crop a image tuple
How make customised dataset for semantic segmentation?
Loss problem in net finetuning
Got cuda out of memory while implementing LeNet5
Data Augmentor for 3D images
Using PyTorch Transformers
Apply same transformation to CycleGAN for same patches
(Oktai Tatanov) #8

what is TF in your code?


#9

Ah sorry, it’s:

import torchvision.transforms.functional as TF

(Kelam Goutam) #10

Assuming both Input and ground truth are images. If we can concatenate input and GT along the axis and then pass the concatenated image through torchvision.transforms.RandomHorizontalFlip() [say]. Then it makes sure that the GT is also flipped when the corresponding input is flipped. I am not sure whether it will work or not practically since I have not tried but theoretically makes sense to me.


#11

The current transformations work with PIL.Images, so that your concatenated image might not be recognized as a valid image. Besides that, it seems to be a good idea.


(Kelam Goutam) #12

Thanks for the clarification. I did Random Horizontal flip by generating the random number and suppose it is greater than a threshold just do a normal horizontal flip of both input and GT.


(Yuqli) #13

Just to add on this thread - the linked PyTorch tutorial on picture loading is kind of confusing. The author does both import skimage import io, transform, and from torchvision import transforms, utils.

For transform, the authors uses a resize() function and put it into a customized Rescale class. For transforms, the author uses the transforms.Compose function to organize two transformations. But they are from two different modules!

To add to the confusion, torchvision transforms also has its own API called Resize() which is the same name with the one in skimage module.


(Preet Khaturia) #14

@ptrblck Where can I get the index from?


#15

Do you mean the index in __getitem__?
You can pass the index directly to your Dataset:

data, target = dataset[0]

or the sampler will do it for you, if you are using a DataLoader.


#16

(Ilyajob05) #17

You can also use the standard functions from the library torchvision
example:

from PIL import Image
from torchvision import transforms
from torchvision.transforms import functional as tvF
...
# create custom class transform
class RRC(transforms.RandomResizedCrop):
    def __call__(self, img1, img2):
        assert img1.size == img2.size
        # fix parameter
        i, j, h, w = self.get_params(img1, self.scale, self.ratio)
        # return the image with the same transformation
        return [tvF.resized_crop(img1, i, j, h, w, self.size, self.interpolation), tvF.resized_crop(img2, i, j, h, w, self.size, self.interpolation)]

imgInput = Image.open('input.png')
imgTarget = Image.open('target.png')   

imgInput, imgTarget = RRC(inputSize, scale=(0.5, 1.0), ratio=(5.0/6.0, 6.0/5.0))(imgInput, imgTarget)