def main():
"""
Convert TXT to PLY.
"""
parser = argparse.ArgumentParser(description='Convert TXT to PLY.')
parser.add_argument('input', type=str, help='TXT file.')
parser.add_argument('output', type=str, help='PLY file.')
args = parser.parse_args()
if not os.path.exists(args.input):
print('Input file does not exist.')
exit(1)
point_cloud = PointCloud.from_txt(args.input)
print('Read %s.' % args.input)
point_cloud.to_ply(args.output)
print('Wrote %s.' % args.output)
print('[Data] read ' +
common.filename(config, 'part_space_file', '_f.h5', dataset))
#statistics = statistics.reshape(1, 1, statistics.shape[0], statistics.shape[1], statistics.shape[2])
#statistics = np.repeat(statistics, space.shape[0], axis=0)
#print(space.shape, statistics.shape)
#invalid_space = space*statistics
points = []
point_dir = common.filename(config, 'bounding_box_txt_directory', '',
dataset) + '/'
for file in os.listdir(point_dir):
point_file = point_dir + file
point_cloud = PointCloud.from_txt(point_file)
#print('[Data] read ' + point_file)
points.append(point_cloud.points.shape[0])
frames = [1] * inputs.shape[0]
frame_dir = common.filename(config, 'velodyne_individual_gt_txt_directory',
'', dataset) + '/'
for i in range(inputs.shape[0]):
for k in range(-config['gt_range'], config['gt_range'] + 1,
config['gt_skip']):
if k == 0:
continue
txt_file = frame_dir + '%d_%d_%d.txt' % (i, k, frames[i])
from point_cloud import PointCloud
if __name__ == '__main__':
if len(sys.argv) < 2:
print('[Data] Usage python 13_ply_observations.py config_folder')
exit(1)
config_folder = sys.argv[1] + '/'
assert os.path.exists(
config_folder), 'directory %s does not exist' % config_folder
config_files = [
config_file for config_file in os.listdir(config_folder)
if not (config_file.find('prior') > 0)
]
for config_file in config_files:
print('[Data] reading ' + config_folder + config_file)
config = utils.read_json(config_folder + config_file)
for k in range(config['n_observations']):
txt_directory = common.dirname(config, 'txt_gt_dir') + str(k) + '/'
assert os.path.exists(txt_directory)
ply_directory = common.dirname(config, 'ply_gt_dir') + str(k) + '/'
if not os.path.exists(ply_directory):
os.makedirs(ply_directory)
for filename in os.listdir(txt_directory):
point_cloud = PointCloud.from_txt(txt_directory + filename)
point_cloud.to_ply(ply_directory + filename[:-4] + '.ply')
print('[Data] wrote ' + ply_directory + filename[:-4] + '.ply')
inputs.shape[3], inputs.shape[4]))
print('[Data] read ' + common.filename(config, 'input_file'))
space = utils.read_hdf5(common.filename(config, 'space_file'))
space = space.reshape(
(space.shape[0] * n_observations, 1, space.shape[2],
space.shape[3], space.shape[4]))
print('[Data] read ' + common.filename(config, 'space_file'))
points = []
point_dir = common.dirname(config, 'txt_gt_dir') + '/'
for k in range(n_observations):
k_point_dir = point_dir + '/%d/' % k
for txt_file in os.listdir(k_point_dir):
point_cloud = PointCloud.from_txt(k_point_dir + txt_file)
points.append(point_cloud.points.shape[0])
points = np.array(points)
occupied = np.squeeze(
np.sum(np.sum(np.sum(targets, axis=4), axis=3), axis=2))
observed_points = np.squeeze(
np.sum(np.sum(np.sum(inputs, axis=4), axis=3), axis=2))
observed_space = np.squeeze(
np.sum(np.sum(np.sum(space, axis=4), axis=3), axis=2))
mask = np.zeros(inputs.shape)
mask[inputs == 1] = 1
mask[space == 1] = 1
observed_total = float(np.sum(mask))
print('[Data] Usage python create_cuboids.py config_folder')
exit(1)
config_file = sys.argv[1]
assert os.path.exists(config_file)
config = utils.read_json(config_file)
dataset = ntpath.basename(config_file)[:-5]
orig_dir = common.filename(config, 'bounding_box_txt_directory', '',
dataset)
gt_dir = common.filename(config, 'velodyne_gt_txt_directory', '', dataset)
gt_files = utils.read_ordered_directory(gt_dir)
for i in range(len(gt_files)):
orig_file = orig_dir + '/' + ntpath.basename(gt_files[i])
orig_point_cloud = PointCloud.from_txt(orig_file)
gt_point_cloud = PointCloud.from_txt(gt_files[i])
print('[Data] read ' + orig_file)
print('[Data] read ' + gt_files[i])
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
xx = gt_point_cloud.points[:, 0]
yy = gt_point_cloud.points[:, 1]
zz = gt_point_cloud.points[:, 2]
ax.scatter(xx, yy, zz, c='b', marker='+', s=25)
xx = orig_point_cloud.points[:, 0]
yy = orig_point_cloud.points[:, 1]
i_ltsdf_inputs[k] = np.min(ltsdf_inputs[perm], axis=0)
# Also fuse the actual point clouds!
txt_directories = utils.read_ordered_directory(
common.dirname(config, 'txt_gt_dir'))
txt_directory = common.dirname(config, 'txt_gt_dir',
i) + '%d/' % k
utils.makedir(txt_directory)
for n in range(N):
point_cloud = PointCloud()
print('[Data] +')
for j in range(perm.shape[0]):
txt_file = txt_directories[perm[j]] + '/%d.txt' % n
print('[Data] | read ' + txt_file)
point_cloud_j = PointCloud.from_txt(txt_file)
point_cloud.points = np.concatenate(
(point_cloud.points, point_cloud_j.points), axis=0)
txt_file = txt_directory + '%d.txt' % n
point_cloud.to_txt(txt_file)
print('[Data] wrote ' + txt_file)
i_inputs = np.swapaxes(i_inputs, 0, 1)
i_space = np.swapaxes(i_space, 0, 1)
i_sdf_inputs = np.swapaxes(i_sdf_inputs, 0, 1)
i_tsdf_inputs = np.swapaxes(i_tsdf_inputs, 0, 1)
i_ltsdf_inputs = np.swapaxes(i_ltsdf_inputs, 0, 1)
print(i_inputs.shape)
print(i_space.shape)
