I am new to Tensorflow and was trying to apply data augmentation to my dataset using the ImageDataGenerator. Turns out that it has various parameters that the tensorflow backend uses to apply data augmentation to each image on the fly while feeding the image to the CNN. But no where in the documentation could I find the number of variations of each image that are generated as well as the total number of images fed to the CNN after the applying data augmentation. Any help would be greatly appreciated.
Number of variants of image produced after applying data augmentation with tensorflow
Viewed 488
1 Answers
When called using .flow() method, the data generator will create one batch size of images for each iteration. This gives the total number of images fed to CNN. The Keras documentation states for FLOW method : " Takes data & label arrays, generates batches of augmented data."
Related