I think your implementation using generators is better than the original one [hao ye], simpler and more concise. It can be, as well, a nice platform for other deep learning techniques.
You used generators to implement data augmentation, and in the same time a good management of the memory, well done man.
may I ask you If you know why the ('channel_train.npy') data is used in the validation generation but the test data was never used in the whole program? thanks in advance!
I think your implementation using generators is better than the original one [hao ye], simpler and more concise. It can be, as well, a nice platform for other deep learning techniques.
You used generators to implement data augmentation, and in the same time a good management of the memory, well done man.