When using Keras 2.0 ImageDataGenerator and in particular the flow_from_directory() method to perform data augmentation, how many images are generated and in what order? Do we have a way to control the number of augmented replicates produced by shifting, rotating, scaling etc from each original?

Let's assume for simplicity that shuffle is turned off and that we have only 3 images A, B, C.

Let's also assume that the ImageDataGenerator 'maps' each original image to 4 (the original and 3 transformed). A first question is how the number 4 is fixed and controlled.

The second question is in what order. Let's give some alternative examples:

1) A, B, C, A1, B1, C1, A2, B2, C2,....

2) A, A1, A2, ..., B, B1, B2, ..., C, C1, C2, C3, C4

3) Some other random way

Your advice will be appreciated.


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.