Question about utils.joint_transforms.Scale()

[RXZ1452132]

Hi~ I find one question in train/cityscapes-fcn/train.py: short_size = int(min(args['input_size']) / 0.875) train_joint_transform = joint_transforms.Compose([ joint_transforms.Scale(short_size), joint_transforms.RandomCrop(args['input_size']), joint_transforms.RandomHorizontallyFlip() ]) And this is the code of function _joint_transforms.Scale(shortsize):

class Scale(object): def init(self, size): self.size = size

def __call__(self, img, mask):
    assert img.size == mask.size
    w, h = img.size
    if (w >= h and w == self.size) or (h >= w and h == self.size):
        return img, mask
    if w > h:
        ow = self.size
        oh = int(self.size * h / w)
        return img.resize((ow, oh), Image.BILINEAR), mask.resize((ow, oh), Image.NEAREST)
    else:
        oh = self.size
        ow = int(self.size * w / h)
        return img.resize((ow, oh), Image.BILINEAR), mask.resize((ow, oh), Image.NEAREST)

for example: the image size is 512256, and args['input_size'] =(256,512) Dose that mean resize the image to 292146(2568/7=292, 256292/512=146)? And after that function joint_transforms.RandomCrop resize it back to 256*512. Why should we do such thing?