I just installed the Transfer Learning Toolkit. Kudos to the NVIDIA team for the great work! I wish to know whether there is a provision to plug in custom python code to enable augmentations on-the-fly during training. I wish to add some augmentations which are not provided in the toolkit. Currently, I am forced to dump all the augmentations along with the original images but I feel that is a naive way to go about things.
Please help me out.