Load my own test data (images)

1

Hi,

I just want to know, how can I load my own test data (image.jpg) in pytorch in order to test my CNN.

I’ve tried test_data = datasets.ImageFolder(data_dir + '/test_cnn', transform=test_transforms)

but during the test, I have too many indices because my test data is automatically labeled by ImageFolder. A solution is to drop the classes but I don’t know How

Thanks for helping

2

An easy solution would be to just drop the labels and use your test data in the DataLoader loop:

for data, _ in loader:
    data = data.to(device)
    ...

This approach is a bit hacky, so the cleaner solution would be to write your own test dataset:

class TestDataset(Dataset):
    def __init__(self, path, transform=None):
        self.image_paths = glob.glob(path + '*.jpg')
        self.transform = transform

    def __getitem__(self, index):
        x = Image.open(self.image_paths[index])
        if self.transform is not None:
            x = self.transform(x)

        return x

    def __len__(self):
        return len(self.image_paths)
1 Like
3

Thanks for your answer. But How can you drop labels? (this is my problem)

4

Sorry I’m a bit beginner in DL area

5

No need to excuse for that. :wink:
I’ve just used an underscore in the data loading loop, to drop the targets:

for data, _ in loader:
    data = data.to(device)
    ...

In this loop only the data is used and no targets are available.
You could also just get the targets and not use them in any way.

2 Likes
6

Thank you so much, it solve my problem