Dataset creation question

[Piotr94]

Hi!

I'd like to make sure if I understand correctly how the training dataset is created. During each epoch from every category three subdirectories are chosen (line 97 in dataset.py) and then for each subdirectory one sequence is chosen. Finally during every epoch only 64*3 sequences are used (64 is the number of categories), is it right?

Best regards Piotr