How do I sample images only from a specific set of classes among all the available classes without using torchvision.datasets.ImageFolder?

1

I came across this answer https://stackoverflow.com/questions/63975130/how-to-get-only-specific-classes-from-pytorchs-fashionmnist-dataset which samples images from a subset of classes using ImageFolder but I want to achieve the same thing using a custom Dataset class, any suggestions on how can i do it?
I have attached my Custom Dataset class for the references.

class CustomDataset(Dataset):
	def __init__(self, root_dir,  transforms=None):
	        self.root_dir = root_dir
		self.transforms = transforms
		self.img_list = sorted(glob.glob(root_dir + "/*/*"))

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

	def __getitem__(self, idx):
		if torch.is_tensor(idx):
		    idx = idx.tolist()	
		img_name = self.img_list[idx]
		image = io.imread(img_name)
		if self.transforms:
		    sample = self.transforms(image)	
		return image
2

You won’t be able to use the same approach, as your custom Dataset doesn’t return any targets and you thus cannot use them to filter out the samples.

3

@ptrblck
what if i change my custom dataset to return the labels in that case?

4

In case you are creating the targets in the __init__ method or are passing them to the Datset, i.e. they are pre-calculated and not lazily computed, you could apply the same approach as in your posted link by filtering out the .targets as well as the corresponding .img_list or by using a Subset.

5

Okay got it . Thanks!!