e0df81789b43a95c32a8aa9e6ce64baba216c5fe,dataset/batch_image.py,BaseImagesBatch,rotate,#BaseImagesBatch#Any#Any#Any#Any#,196

Before Change

            (always True for images as arrays, can be False for PIL.Images)
        

        return self._rotate_image(self.get(ix, components), angle, preserve_shape, **kwargs)

    @action
    @inbatch_parallel(init="indices", post="assemble")
    def random_rotate(self, ix, components="images", p=1., angle=None, **kwargs):

After Change

            angle"s range to sample from in the form of (min_angle, max_angle), in degrees.
            if float is passed than determenistic rotation will be perfomed
        
        if p > 1 or p < 0:
            raise ValueError("probability must be in [0,1]")

        image = self.get(ix, components)
        if np.random.random() < p:
            angle = angle or (-45., 45.)
            if not isinstance(angle, (float, int)):
                angle = np.random.uniform(*(angle))
            preserve_shape = kwargs.pop("preserve_shape", True)
            return self._rotate_image(image, angle, preserve_shape=preserve_shape, **kwargs)
        return image


    @action

Instances