Increases size of training set and it's annotation data by applying operations (rotation, contrast, color space, etc.) based on selection