Data loader for unlabelled images

Hi, this might be a very silly question but i’m new to fastai and was reading the book. I learned how to create image dataloaders and fine tuned a resnet on it. But i want to test it on test images which are unlabelled. How to create dataloaders for them ?

With DataLoaders.test_dl. Assuming you have a dataloader that takes a list of files and test_files is the list of files to test, you’d do this:

test_dl = learn.dls.test_dl(test_files)
preds, _ = learn.get_preds(dl=test_dl)

If your dataloader read pandas dataframes, you’d pass test_dl a dataframe of test items.

For more details you can look at my inference tutorial here.

1 Like

Thanks a lot…