def meshlab_poisson(pointcloud):
r''' Runs the meshlab ball pivoting algorithm.
Args:
pointcloud (numpy tensor): input point cloud
'''
with tempfile.TemporaryDirectory() as tmpdir:
script_path = os.path.join(tmpdir, 'script.mlx')
input_path = os.path.join(tmpdir, 'input.ply')
output_path = os.path.join(tmpdir, 'out.off')
# Write script
with open(script_path, 'w') as f:
f.write(FILTER_SCRIPT_RECONSTRUCTION)
# Write pointcloud
export_pointcloud(pointcloud, input_path, as_text=False)
# Export
env = os.environ
subprocess.Popen(
'meshlabserver -i ' + input_path + ' -o ' + output_path + ' -s ' +
script_path,
env=env,
cwd=os.getcwd(),
shell=True,
stdout=subprocess.DEVNULL,
stderr=subprocess.DEVNULL,
).wait()
mesh = trimesh.load(output_path, process=False)
return mesh
def __call__(self, inputs):
self.model.eval()
with torch.no_grad():
points = self.model(inputs)
export_pointcloud(
points.squeeze(0).cpu().numpy(),
'/is/sg/lmescheder/Desktop/points.ply')
batch_size = points.size(0)
T = points.size(1)
# Subsample points if necessary
if T != self.pointcloud_n:
idx = torch.randint(low=0,
high=T,
size=(batch_size, self.pointcloud_n),
device=self.device)
idx = idx[:, :, None].expand(batch_size, self.pointcloud_n, 3)
points = torch.gather(points, dim=1, index=idx)
return points
def export(self, points, pointcloud_dir, modelname, start_idx=0):
''' Saves a sequence of point clouds.
Args:
points (numpy array): point cloud points
pointcloud_dir (str): pointcloud directory
modelname (str): model name
'''
model_folder = os.path.join(pointcloud_dir, modelname,
'%05d' % start_idx)
if not os.path.isdir(model_folder):
os.makedirs(model_folder)
out_files = []
for i, pc in enumerate(points):
pc_out_file = os.path.join(
model_folder,
'%s_%04d_%04d.ply' % (modelname, start_idx, i + 1))
export_pointcloud(pc, pc_out_file)
out_files.append(pc_out_file)
return out_files
except TypeError:
mesh, stats_dict = out, {}
time_dict.update(stats_dict)
# Write output
mesh_out_file = os.path.join(mesh_dir, "%s.off" % modelname)
mesh.export(mesh_out_file)
out_file_dict["mesh"] = mesh_out_file
if generate_pointcloud:
t0 = time.time()
pointcloud = generator.generate_pointcloud(data)
time_dict["pcl"] = time.time() - t0
pointcloud_out_file = os.path.join(pointcloud_dir,
"%s.ply" % modelname)
export_pointcloud(pointcloud, pointcloud_out_file)
out_file_dict["pointcloud"] = pointcloud_out_file
if cfg["generation"]["copy_input"]:
# Save inputs
if input_type == "img":
inputs_path = os.path.join(in_dir, "%s.jpg" % modelname)
inputs = tf.squeeze(data["inputs"], axis=0)
visualize_data(inputs, "img", inputs_path)
out_file_dict["in"] = inputs_path
elif input_type == "voxels":
inputs_path = os.path.join(in_dir, "%s.off" % modelname)
inputs = tf.squeeze(data["inputs"], axis=0)
voxel_mesh = VoxelGrid(inputs).to_mesh()
voxel_mesh.export(inputs_path)
out_file_dict["in"] = inputs_path
if c_it < vis_n_outputs:
# Save inputs
if cfg['generation']['copy_input']:
if input_type == 'pcl_seq':
inputs = data['inputs'][0]
L = inputs.shape[0]
inputs_base_path = os.path.join(in_dir, modelname,
'%04d' % start_idx)
if not os.path.exists(inputs_base_path):
os.makedirs(inputs_base_path)
inputs_path = [
os.path.join(inputs_base_path, '%04d.ply' % i)
for i in range(L)
]
for i in range(L):
export_pointcloud(inputs[i].cpu().numpy(), inputs_path[i],
False)
out_file_dict['in_%d' % i] = inputs_path[i]
# Save output files
for k, filepath in out_file_dict.items():
ext = os.path.splitext(filepath)[1]
out_file = os.path.join(generation_vis_dir,
'%02d_%s%s' % (c_it, k, ext))
shutil.copyfile(filepath, out_file)
model_counter[category_id] += 1
# Create pandas dataframe and save
time_df = pd.DataFrame(time_dicts)
time_df.set_index(['idx'], inplace=True)
time_df.to_pickle(out_time_file)
