My data was shuffled randomly first then I applied a five crop data augmentation. Now my batch went from [8, 3, 256, 256] to [40, 3, 256, 256] as you can see each 256x256 3 channel images were split into 5 crops increasing the total size to 40. I also double checked the array for ordering and the new cropped images are in order form so I had to do the same thing for my labels. A one-hot encoding example would be [1,1,1,1,1, 0,0,0,0,0, 1,1,1,1,1 ... 1,1,1,1,1]. Right now my labels is a [40, 1] vector. My current concern is should I reshuffle my images again? The neural network should recognize each image as an individual image but I am not confident about the effect of the paired ordering labels.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.