Dataset index problem

1

Recently, I’ve written a custom Dataset class to load my dataset. Here is what it looks like:

30
30626×870 25.6 KB

The folder name is 001 and the picture names are from 0 to 5. Here is my mathDataset:

class mathdataset(Dataset):
    def __init__(self, root_dir, transform=None):
        self.root_dir=root_dir
        self.transform=transform
    
    def __len__(self):
        filenames=glob.glob(str(self.root_dir)+'/*.png')
        return len(filenames)
    def __getitem__(self,i):
        img_name=str(self.root_dir)+'/'+str(i)+'.png'
        im=Image.open(img_name)
        image=im.convert('RGB')
        if self.transform:
            image = self.transform(image)
        return image

I’ve written the __len__ and __getitem__ as the doc suggested.

The problem came when I wanted to iterate through the dataset like this:

28
282068×1110 266 KB

Then I checked the len(dataset)
02
It was correct. And the last tensor in the dataset which size is torch.Size([3, 224, 262]) correctly matched the picture 5.png
So is there anything I can do like modify the mathDataset or rename the pictures to get the right output?

2

From the python docs

Note: for loops expect that an IndexError will be raised for illegal indexes to allow proper detection of the end of the sequence.

So how about catching the file not found error and raising an index error instead.

try:
    im=Image.open(img_name)
except FileNotFoundError:
    raise IndexError
3

Thank you, you are correct.
You really help me a lot!