When we use Transforms for data augmentation, does the size of the dataset increases so that to contain both original and transformed images, or is it that just the original images are transformed and the size of dataset remains same?
Augmented image is a new PIL Image object.
def _new(self, im): new = Image() new.im = im new.mode = im.mode new.size = im.size if self.palette: new.palette = self.palette.copy() if im.mode == "P" and not new.palette: from PIL import ImagePalette new.palette = ImagePalette.ImagePalette() new.info = self.info.copy() return new
@Hyungjoo_Cho so, how to make the data number change? such as, I need to make the origin image and the flipped image both into the training data. In the other words, I need to double the origin data. So, how to accomplish this?