def sample_batch_in_all_files(self, batch_size, augment=True):
batch_data = []
batch_label = []
batch_weights = []
for _ in range(batch_size):
points, labels, colors, weights = self.sample_in_all_files(
is_training=True)
if self.use_color:
batch_data.append(np.hstack((points, colors)))
else:
batch_data.append(points)
batch_label.append(labels)
batch_weights.append(weights)
batch_data = np.array(batch_data)
batch_label = np.array(batch_label)
batch_weights = np.array(batch_weights)
if augment:
if self.use_color:
batch_data = provider.rotate_feature_point_cloud(batch_data, 3)
else:
batch_data = provider.rotate_point_cloud(batch_data)
return batch_data, batch_label, batch_weights
def _augment_batch_data(self, batch_data):
rotated_data = rotate_point_cloud(batch_data)
rotated_data = rotate_perturbation_point_cloud(rotated_data)
jittered_data = random_scale_point_cloud(rotated_data[:, :, 0:3])
jittered_data = shift_point_cloud(jittered_data)
jittered_data = jitter_point_cloud(jittered_data)
rotated_data[:, :, 0:3] = jittered_data
return shuffle_points(rotated_data)