Why we use random cropping for the image in deep learning, is it one of data augmentation? What if an important portion of the image is cropped?
Yes, it’s a data augmentation method.
Often your crop size is not much smaller than the original image size, e.g. you are cropping
224x224 patches out of a
However, some techniques (IIRC in Inception) use larger differences between the crop size and original size.