Giving more than one path to Dataloader

vision
1

I have multiple path with folders containing images, is it possible to use pytorch Dataloader reading from all of them?

2

Yes it’s definitely possible! Try searching Google or the docs next time;

You are looking for torchvision.datasets.ImageFolder()!

3

Tnx for your response, I have multiple sources for training. My data folders are like below:

Train/
         class1/
                   pos/
                   neg/

          class2/
                   pos/
                   neg/

When I’m using this with datasets.ImageFolder() it gives wrong labels to images, for example all negative should be 0 and all positives should be 1. how can I do this?