CODEWITHSUNDEEP
pythonpytorch
Guy Barash
Guy Barash

Reputation: 490

removing sample from a Pytorch dataset object

I'm loading images to pytorch using:

train_data = torchvision.datasets.ImageFolder(train_dir, transform=transform)

Which has some samples I want to omit.

Can i remove certain images from the loader? (by index? by name?) Thx

Upvotes: 1

Views: 3538

Answers (1)

Gil Pinsky
Gil Pinsky

Reputation: 2493

You can modify the imgs attribute of an ImageFolder as follows:

train_data.imgs.remove((PATH, INDEX))

where you should replace PATH with the path to the image you wish to remove and INDEX should be replaced with the respective class index (starting from 0 to the number of classes you have). In case you are not sure about the INDEX you can print train_data.imgs to obtain the list of all image paths with their classes in tuples.

Upvotes: 1

Related Questions