Closed Zhaoyu321 closed 2 years ago
They are just the metafiles that define the dataset splits. For example, train1
includes all the images in the training set, while train2
includes all the images in the validation set (or subtrain set) for the evolution search. train1+train2
is used for fine-tuning.
You could also split the training set to train1
and train2
yourself.
I understand, thank you very much for your reply!
Hello! How to get the files under these two paths (--metaA_path --metaB_path) when I train a supernet. In other words, how do I get the files under these two paths by training on my own dataset. Thanks!