Method __call__

dataloader/data_augmentation.py:204–227

Randomly jitter points. jittering is per point. Input: BxNx3 array, original batch of point clouds Return: BxNx3 array, jittered batch of point clouds

(self, e)

        self.p = p

    def __call__(self, e):
        """ Randomly jitter points. jittering is per point.
            Input:
              BxNx3 array, original batch of point clouds
            Return:
              BxNx3 array, jittered batch of point clouds
        """

        sample_shape = (e.shape[0],)
        if self.p < 1.:
            # Create a mask for points to jitter
            m = torch.distributions.categorical.Categorical(probs=torch.tensor([1 - self.p, self.p]))
            mask = m.sample(sample_shape=sample_shape)
        else:
            mask = torch.ones(sample_shape, dtype=torch.int64 )

        mask = mask == 1
        jitter = self.sigma * torch.randn_like(e[mask])

        if self.clip is not None:
            jitter = torch.clamp(jitter, min=-self.clip, max=self.clip)

        e[mask] = e[mask] + jitter
        return e


class RemoveRandomPoints: