Based on my understanding, the transformer randomly modifies the existing data.
Is there any fast ai API that would allow me to expand the training data 10 times by applying random augmentation transformation?
Based on my understanding, the transformer randomly modifies the existing data.
Is there any fast ai API that would allow me to expand the training data 10 times by applying random augmentation transformation?
Yes. Read the docs (vision, transformer section). You can do what you require, and much more. If you want to generate a new dataset rather than apply the transformations every time, take Image.data
and convert to numpy. Then save it with pillow, matplotlib, or opencv.
Do you mean vision.transform? I checked it out but it doesn’t mention anything about expanding train data set. Could you specify which document you’re referring to?