How to apply it to my own dataset？

[ghost]

I tried to apply it to my own dataset, but it didn't work because unmarked data had to be enhanced twice. Thank you very much for who can help me

[sakumashirayuki]

what is your problem? I suggest write your own dataset class to read your images. Here is the dataset class read images from a path

class YourDatasetName(Dataset):
    def __init__(self, root_dir, transform=None):
        self.root_dir = root_dir
        self.transform = transform
        self.image_lists = os.listdir(root_dir)

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

    def __getitem__(self, item):
        image_path = os.path.join(self.root_dir, self.image_lists[item])
        image = Image.open(image_path)
        image = image.convert('RGB')
        if self.transform:
            img_tensor_1 = self.transform(image)
            img_tensor_2 = self.transform(image)
        return img_tensor_1, img_tensor_2

[ghost]

Thank you for your help. Under your inspiration, I solved this problem. Thank you very much

[thiyagu145]

@sakumashirayuki Have you tried training in a multi gpu setting with DP/DDP?