How to implement image augmentation in the model. do we need to do image augmentation during the creation of the dataset??Also does image augmentation increase the number of images or not??
(padding the reply to > 20 characters)
how to add the step of image augmentation in creating our dataset…Ive created a dataset having array of images…So do I need to pass it through the keras augmentation function?? How to know the number of transformation each image gets??
There are instructions in the notebook, if you are talking about the Transfer Learning with MobilNet assignment. They walk you through what to do. Have you read them carefully?
The other high level thing to note is that the augmentation, as they implement it in that case, happens “on the fly”, meaning that the augmented images are not saved in the dataset. So every time you run the training you get randomly different inputs even though they are all based on the input dataset.
I want to do it in the normal CNN done in week1
Ok, you have a worked example to start from, but that is in TF of course.