Short excerpt from a YouTube video ("Augmenting Data (TensorFlow in Practice)") discussing practical techniques for image data augmentation in TensorFlow. Speakers explain using an image data generator so augmentations (warping, skewing, etc.) occur in memory as images are loaded into the neural network, avoiding overwriting on-disk files and preventing large increases in storage requirements.
But there's an important trick to how you do this in TensorFlow as well. To not take an image, warp it, skew it, and then blow up the memory requirements. So TensorFlow makes it really easy to do this?
Yes. So you will learn a lot about the image generator and the image data generator, where the idea is that you're not going to edit the images directly on the drive as they get flowed off the directory. Then the augmentation will take place in memory as they're being loaded into the neural network for training. So if you're dealing with a data set and you want to experiment with different augmentations, you're not overriding your data.
