def _pre_transform_fragments(self):
"""
apply pre_transform on fragments (ply) and save the results
"""
out_dir = osp.join(self.processed_dir, 'test', 'fragment')
if files_exist([out_dir]): # pragma: no cover
return
makedirs(out_dir)
# table to map fragment numper with
self.table = dict()
list_scene = [f for f in os.listdir(osp.join(self.raw_dir, "test"))]
for scene_path in list_scene:
fragment_dir = osp.join(self.raw_dir, "test", scene_path)
if (osp.isfile(fragment_dir)):
continue
list_fragment_path = sorted(
[f for f in os.listdir(fragment_dir) if 'pcd' in f])
for i, f_p in enumerate(list_fragment_path):
fragment_path = osp.join(fragment_dir, f_p)
out_dir = osp.join(self.processed_dir, "test", 'fragment',
scene_path)
makedirs(out_dir)
out_path = osp.join(out_dir,
'fragment_{:06d}.pt'.format(find_int(f_p)))
pos = torch.from_numpy(BasePCRBTest.read_pcd(fragment_path)[0])
data = Data(pos=pos)
if (self.pre_transform is not None):
data = self.pre_transform(data)
torch.save(data, out_path)
def _compute_matches_between_fragments(self):
ind = 0
out_dir = osp.join(self.processed_dir, "test", "matches")
if files_exist([out_dir]): # pragma: no cover
return
makedirs(out_dir)
list_scene = os.listdir(osp.join(self.raw_dir, "test"))
for scene in list_scene:
if (osp.isfile(osp.join(self.raw_dir, "test", scene))):
continue
path_log = osp.join(self.raw_dir, "test", scene + "_global.txt")
list_pair = BasePCRBTest.parse_pair_files(path_log)
for i, pair in enumerate(list_pair):
path1 = osp.join(
self.processed_dir, "test", 'fragment', scene,
'fragment_{:06d}.pt'.format(find_int(pair["source_name"])))
path2 = osp.join(
self.processed_dir, "test", 'fragment', scene,
'fragment_{:06d}.pt'.format(find_int(pair["target_name"])))
data1 = torch.load(path1)
data2 = torch.load(path2)
match = compute_overlap_and_matches(data1, data2,
self.max_dist_overlap)
match['path_source'] = path1
match['path_target'] = path2
match['name_source'] = pair["source_name"]
match['name_target'] = pair["target_name"]
match['scene'] = scene
match['trans'] = pair["trans"]
out_path = osp.join(self.processed_dir, "test", 'matches',
'matches{:06d}.npy'.format(ind))
np.save(out_path, match)
ind += 1
def _pre_transform_fragment(self, mod):
"""
pre_transform raw fragments and save it into fragments
"""
out_dir = osp.join(self.processed_dir, mod, 'fragment')
if files_exist([out_dir]): # pragma: no cover
return
makedirs(out_dir)
for scene_path in os.listdir(osp.join(self.raw_dir, mod)):
# TODO list the right sequences.
list_seq = [
f for f in os.listdir(osp.join(self.raw_dir, mod, scene_path))
if 'seq' in f
]
for seq in list_seq:
in_dir = osp.join(self.processed_dir, mod, 'raw_fragment',
scene_path, seq)
out_dir = osp.join(self.processed_dir, mod, 'fragment',
scene_path, seq)
makedirs(out_dir)
list_fragment_path = sorted(
[f for f in os.listdir(in_dir) if 'fragment' in f])
for path in list_fragment_path:
data = torch.load(osp.join(in_dir, path))
if (self.pre_transform is not None):
data = self.pre_transform(data)
torch.save(data, osp.join(out_dir, path))
def download(self):
folder = osp.join(self.raw_dir, "test")
print(folder)
if files_exist([folder]): # pragma: no cover
log.warning("already downloaded {}".format("test"))
return
else:
makedirs(folder)
log.info("Download elements in the file {}...".format(folder))
for name, url in self.DATASETS:
req = requests.get(url)
with open(osp.join(folder, name + ".zip"), "wb") as archive:
archive.write(req.content)
with ZipFile(osp.join(folder, name + ".zip"), "r") as zip_obj:
log.info("extracting dataset {}".format(name))
zip_obj.extractall(osp.join(folder, name))
log.info("converting to PCD")
asl_to_pcd(osp.join(folder, name))
file_not_to_remove = glob.glob(osp.join(folder, name, "*.pcd"))
filelist = glob.glob(osp.join(folder, name, "*"))
for file_to_remove in filelist:
if file_to_remove not in file_not_to_remove:
os.remove(file_to_remove)
os.remove(osp.join(folder, name + ".zip"))
self.download_pairs(folder)
def _compute_matches_between_fragments(self):
ind = 0
out_dir = osp.join(self.processed_dir, "test", "matches")
if files_exist([out_dir]): # pragma: no cover
return
makedirs(out_dir)
list_scene = os.listdir(osp.join(self.raw_dir, "test"))
for scene in list_scene:
path_log = osp.join(self.raw_dir, "test", scene, "gt.log")
list_pair_num, list_mat = read_gt_log(path_log)
for i, pair in enumerate(list_pair_num):
path1 = osp.join(self.processed_dir, "test", 'fragment', scene,
'fragment_{:06d}.pt'.format(pair[0]))
path2 = osp.join(self.processed_dir, "test", 'fragment', scene,
'fragment_{:06d}.pt'.format(pair[1]))
data1 = torch.load(path1)
data2 = torch.load(path2)
match = compute_overlap_and_matches(
data1,
data2,
self.max_dist_overlap,
trans_gt=torch.from_numpy(np.linalg.inv(list_mat[i])).to(
data1.pos.dtype))
match['path_source'] = path1
match['path_target'] = path2
match['name_source'] = str(pair[0])
match['name_target'] = str(pair[1])
match['scene'] = scene
out_path = osp.join(self.processed_dir, "test", 'matches',
'matches{:06d}.npy'.format(ind))
np.save(out_path, match)
ind += 1
def _create_fragment(self, mod):
r"""
create fragments from the rgbd frames ie a partial reconstruction of
the scene with some frames(usually 50).
We will only use the first sequence for each scene
"""
print("Create fragment from RGBD frames...")
if files_exist([osp.join(self.processed_dir, mod,
'raw_fragment')]): # pragma: no cover
log.warning("the fragments on mode {} already exists".format(mod))
return
for scene_path in os.listdir(osp.join(self.raw_dir, mod)):
# TODO list the right sequences.
list_seq = [
f for f in os.listdir(osp.join(self.raw_dir, mod, scene_path))
if 'seq' in f
]
for seq in list_seq:
frames_dir = osp.join(self.raw_dir, self.mode, scene_path, seq)
out_dir = osp.join(self.processed_dir, mod, 'raw_fragment',
scene_path, seq)
makedirs(out_dir)
path_intrinsic = osp.join(self.raw_dir, self.mode, scene_path,
'camera-intrinsics.txt')
list_path_frames = sorted([
osp.join(frames_dir, f) for f in os.listdir(frames_dir)
if 'png' in f and 'depth' in f
])
# list_path_color = sorted([osp.join(frames_dir, f)
# for f in os.listdir(frames_dir)
# if 'png' in f and 'color' in f])
list_path_trans = sorted([
osp.join(frames_dir, f) for f in os.listdir(frames_dir)
if 'pose' in f and 'txt' in f
])
# compute each fragment and save it
if (not self.is_fine):
rgbd2fragment_rough(list_path_frames,
path_intrinsic,
list_path_trans,
out_dir,
self.num_frame_per_fragment,
pre_transform=None)
else:
assert len(list_path_frames) == len(list_path_trans), \
log.error("For the sequence {},"
"the number of frame "
"and the number of "
"pose is different".format(frames_dir))
rgbd2fragment_fine(list_path_frames,
path_intrinsic,
list_path_trans,
out_dir,
self.num_frame_per_fragment,
voxel_size=self.tsdf_voxel_size,
pre_transform=None,
depth_thresh=self.depth_thresh,
limit_size=self.limit_size)
def download(self):
# we download the raw RGBD file for the train and the validation data
folder = osp.join(self.raw_dir, self.mode)
if files_exist([folder]): # pragma: no cover
log.warning("already downloaded {}".format(self.mode))
return
log.info("Download elements in the file {}...".format(folder))
for url in self.dict_urls[self.mode]:
path = download_url(url, folder, self.verbose)
extract_zip(path, folder, self.verbose)
os.unlink(path)
def _pre_transform_fragments(self):
"""
apply pre_transform on fragments (ply) and save the results
"""
out_dir = osp.join(self.processed_dir, "test", "fragment")
if files_exist([out_dir]): # pragma: no cover
return
makedirs(out_dir)
# table to map fragment numper with
self.table = dict()
list_scene = [f for f in os.listdir(osp.join(self.raw_dir, "test"))]
for scene_path in list_scene:
pose_path = osp.join(self.raw_dir, "test",
"pose_{}.csv".format(scene_path))
fragment_dir = osp.join(self.raw_dir, "test", scene_path)
if osp.isfile(fragment_dir):
continue
list_fragment_path = sorted(
[f for f in os.listdir(fragment_dir) if "pcd" in f])
for i, f_p in enumerate(list_fragment_path):
fragment_path = osp.join(fragment_dir, f_p)
out_dir = osp.join(self.processed_dir, "test", "fragment",
scene_path)
makedirs(out_dir)
if (self.is_name_path_int):
out_path = osp.join(
out_dir, "fragment_{:06d}.pt".format(find_int(f_p)))
else:
out_path = osp.join(out_dir, "{}.pt".format(f_p[:-4]))
pos = torch.from_numpy(
BasePCRBTest.read_pcd(fragment_path)[0][:, :3]).float()
data = Data(pos=pos)
if self.pre_transform is not None:
data = self.pre_transform(data)
if (osp.exists(pose_path)):
ind = find_int(f_p)
df = pd.read_csv(pose_path)
center = torch.tensor(
[[df[' T03'][ind], df[' T13'][ind],
df[' T23'][ind]]]).float().unsqueeze(0)
ind_sensors, _ = dense_knn(data.pos.unsqueeze(0).float(),
center,
k=1)
data.ind_sensors = ind_sensors[0][0]
else:
log.warn("No censors data")
torch.save(data, out_path)
def _pre_transform_fragments_ply(self):
"""
apply pre_transform on fragments (ply) and save the results
"""
out_dir = osp.join(self.processed_dir, "test", "fragment")
if files_exist([out_dir]): # pragma: no cover
return
makedirs(out_dir)
ind = 0
# table to map fragment numper with
self.table = dict()
for scene_path in os.listdir(osp.join(self.raw_dir, "test")):
fragment_dir = osp.join(self.raw_dir, "test", scene_path)
list_fragment_path = sorted(
[f for f in os.listdir(fragment_dir) if "ply" in f])
for i, f_p in enumerate(list_fragment_path):
fragment_path = osp.join(fragment_dir, f_p)
out_dir = osp.join(self.processed_dir, "test", "fragment",
scene_path)
makedirs(out_dir)
out_path = osp.join(out_dir,
"fragment_{:06d}.pt".format(find_int(f_p)))
# read ply file
with open(fragment_path, "rb") as f:
data = PlyData.read(f)
pos = [
torch.tensor(data["vertex"][axis])
for axis in ["x", "y", "z"]
]
pos = torch.stack(pos, dim=-1)
data = Data(pos=pos)
if self.pre_transform is not None:
data = self.pre_transform(data)
torch.save(data, out_path)
self.table[ind] = {
"in_path": fragment_path,
"scene_path": scene_path,
"fragment_name": f_p,
"out_path": out_path,
}
ind += 1
# save this file into json
with open(osp.join(out_dir, "table.json"), "w") as f:
json.dump(self.table, f)
def _pre_transform_fragments_ply(self):
"""
apply pre_transform on fragments (ply) and save the results
"""
out_dir = osp.join(self.processed_dir, 'test', 'fragment')
if files_exist([out_dir]): # pragma: no cover
return
makedirs(out_dir)
ind = 0
# table to map fragment numper with
self.table = dict()
for scene_path in os.listdir(osp.join(self.raw_dir, "test")):
fragment_dir = osp.join(self.raw_dir, "test", scene_path)
list_fragment_path = sorted(
[f for f in os.listdir(fragment_dir) if 'ply' in f])
for i, f_p in enumerate(list_fragment_path):
fragment_path = osp.join(fragment_dir, f_p)
out_dir = osp.join(self.processed_dir, "test", 'fragment',
scene_path)
makedirs(out_dir)
out_path = osp.join(out_dir,
'fragment_{:06d}.pt'.format(find_int(f_p)))
# read ply file
with open(fragment_path, 'rb') as f:
data = PlyData.read(f)
pos = ([
torch.tensor(data['vertex'][axis])
for axis in ['x', 'y', 'z']
])
pos = torch.stack(pos, dim=-1)
data = Data(pos=pos)
if (self.pre_transform is not None):
data = self.pre_transform(data)
torch.save(data, out_path)
self.table[ind] = {
'in_path': fragment_path,
'scene_path': scene_path,
'fragment_name': f_p,
'out_path': out_path
}
ind += 1
# save this file into json
with open(osp.join(out_dir, 'table.json'), 'w') as f:
json.dump(self.table, f)
def _save_patches(self, mod):
"""
save patch to load it offline for the training
"""
p_extractor = PatchExtractor(self.radius_patch)
out_dir = osp.join(self.processed_dir, mod, 'patches')
if files_exist([out_dir]): # pragma: no cover
return
makedirs(out_dir)
match_dir = osp.join(self.processed_dir, mod, 'matches')
idx = 0
for i in range(len(os.listdir(match_dir))):
match = np.load(osp.join(match_dir, 'matches{:06d}.npy'.format(i)),
allow_pickle=True).item()
for _ in range(self.num_random_pt):
data_source = torch.load(match['path_source'])
data_target = torch.load(match['path_target'])
rand = np.random.randint(0, len(match['pair']))
data_source = p_extractor(data_source, match['pair'][rand][0])
data_target = p_extractor(data_target, match['pair'][rand][1])
if (self.pre_transform_patch is not None):
data_source = self.pre_transform_patch(data_source)
data_target = self.pre_transform_patch(data_target)
if (self.pre_filter is not None):
if (self.pre_filter(data_source)
and self.pre_filter(data_target)):
torch.save(
data_source,
osp.join(out_dir,
'patches_source{:06d}.pt'.format(idx)))
torch.save(
data_target,
osp.join(out_dir,
'patches_target{:06d}.pt'.format(idx)))
idx += 1
else:
torch.save(
data_source,
osp.join(out_dir,
'patches_source{:06d}.pt'.format(idx)))
torch.save(
data_target,
osp.join(out_dir,
'patches_target{:06d}.pt'.format(idx)))
idx += 1
def download(self):
folder = osp.join(self.raw_dir, "test")
if files_exist([folder]): # pragma: no cover
log.warning("already downloaded {}".format("test"))
return
else:
makedirs(folder)
ftp = FTP('asrl3.utias.utoronto.ca')
ftp.login()
log.info("Download elements in the file {}...".format(folder))
for name, url in self.DATASETS:
zip_file = osp.join(folder, name + '.zip')
log.info("Downloading dataset %s" % name)
ftp.retrbinary('RETR ' + url, open(zip_file, 'wb').write)
with ZipFile(zip_file, 'r') as zip_obj:
log.info("Extracting dataset %s" % name)
zip_obj.extractall(folder)
with os.scandir(osp.join(folder, name)) as directory:
log.info("Configuring dataset %s" % name)
for entry in directory:
if entry.is_dir():
base_path = entry.path + "/" + entry.name
file_name = base_path + ".xyz"
ground_truth_name = base_path + ".gt"
pcd_file_name = entry.path + ".pcd"
pcd = open3d.io.read_point_cloud(
file_name,
format="xyz",
remove_nan_points=True,
remove_infinite_points=True,
print_progress=False)
ground_truth = numpy.loadtxt(ground_truth_name)
pcd.transform(ground_truth)
open3d.io.write_point_cloud(pcd_file_name,
pcd,
write_ascii=True,
compressed=False,
print_progress=False)
shutil.rmtree(entry.path)
os.remove(zip_file)
gdown.download(
"https://drive.google.com/uc?id=1marTTFGjlDTb-MLj7pm5zV1u-0IS-xFc",
folder + "/p2at_met/box_map.pcd",
quiet=True)
self.download_pairs(folder)
def download(self):
folder = os.path.join(self.raw_dir, "test")
if files_exist([folder]): # pragma: no cover
log.warning("already downloaded {}".format("test"))
return
else:
makedirs(folder)
log.info("Download elements in the file {}...".format(folder))
for name, url in self.DATASETS:
log.info(f'Downloading sequence {name}')
filename = os.path.join(folder, name + ".zip")
gdown.download(url, filename, quiet=False)
with ZipFile(filename, 'r') as zip_obj:
zip_obj.extractall(folder)
os.remove(filename)
self.download_pairs(folder)
def _pre_transform_fragment(self, mod):
"""
read raw fragment, rotate the raw fragment using the calibration and the given pose,
pre transform raw_fragment and save it into fragments
"""
if files_exist([osp.join(self.processed_dir, mod,
"fragment")]): # pragma: no cover
return
in_dir = osp.join(self.raw_dir, "dataset")
list_drive = self.dict_seq[mod]
for drive in list_drive:
out_dir = osp.join(self.processed_dir, mod, "fragment",
"{:02d}".format(drive))
makedirs(out_dir)
path_frames = osp.join(in_dir, "sequences", "{:02d}".format(drive),
"velodyne")
T_calib = read_calib_file(
osp.join(in_dir, "sequences", "{:02d}".format(drive),
"calib.txt"))
all_poses = np.genfromtxt(
osp.join(in_dir, "refined_poses", "{:02d}.txt".format(drive)))
list_name_frames = sorted(
[f for f in os.listdir(path_frames) if "bin" in f])
for i, name in enumerate(list_name_frames):
pose = all_poses[i].reshape((3, 4))
xyzr = np.fromfile(osp.join(path_frames, name),
dtype=np.float32).reshape((-1, 4))
xyzr[:, :3] = xyzr[:, :3].dot(T_calib[:3, :3].T) + T_calib[:3,
3]
xyzr[:, :3] = xyzr[:, :3].dot(pose[:3, :3].T) + pose[:3, 3]
# store position of the car to filter some frames
pos_sensor = pose[:3, :3].dot(T_calib[:3, 3]) + pose[:3, 3]
data = Data(
pos=torch.from_numpy(xyzr[:, :3]),
reflectance=torch.from_numpy(xyzr[:, 3]),
pos_sensor=torch.from_numpy(pos_sensor),
)
out_path = osp.join(out_dir, name.split(".")[0] + ".pt")
if self.pre_transform is not None:
data = self.pre_transform(data)
torch.save(data, out_path)
def download(self):
"""
TODO: download the datasets
"""
folder = osp.join(self.raw_dir, "test")
print(folder)
if files_exist([folder]): # pragma: no cover
log.warning("already downloaded {}".format("test"))
return
else:
makedirs(folder)
log.info("Download elements in the file {}...".format(folder))
req = requests.get(self.url)
with open(osp.join(folder, self.name + ".zip"), "wb") as archive:
archive.write(req.content)
with ZipFile(osp.join(folder, self.name + ".zip"), "r") as zip_obj:
log.info("extracting dataset {}".format(self.name))
zip_obj.extractall(folder)
def download(self):
folder_test = osp.join(self.raw_dir, "test")
if files_exist([folder_test]): # pragma: no cover
log.warning("already downloaded {}".format("test"))
return
for url_raw in self.list_urls_test:
url = url_raw.split("\n")[0]
path = download_url(url, folder_test)
extract_zip(path, folder_test)
log.info(path)
folder = path.split(".zip")[0]
os.unlink(path)
path_eval = download_url(url.split(".zip")[0] + "-evaluation.zip", folder)
extract_zip(path_eval, folder)
os.unlink(path_eval)
folder_eval = path_eval.split(".zip")[0]
for f in os.listdir(folder_eval):
os.rename(osp.join(folder_eval, f), osp.join(folder, f))
shutil.rmtree(folder_eval)
def _compute_matches_between_fragments(self, mod):
out_dir = osp.join(self.processed_dir, mod, 'matches')
if files_exist([out_dir]): # pragma: no cover
return
makedirs(out_dir)
for scene_path in os.listdir(osp.join(self.raw_dir, mod)):
list_seq = sorted([
f for f in os.listdir(osp.join(self.raw_dir, mod, scene_path))
if 'seq' in f
])
for seq in list_seq:
log.info("{}, {}".format(scene_path, seq))
fragment_dir = osp.join(self.processed_dir, mod, 'fragment',
scene_path, seq)
list_fragment_path = sorted([
osp.join(fragment_dir, f) for f in os.listdir(fragment_dir)
if 'fragment' in f
])
log.info("compute_overlap_and_matches")
ind = 0
for path1 in list_fragment_path:
for path2 in list_fragment_path:
if path1 < path2:
out_path = osp.join(
out_dir, 'matches{:06d}.npy'.format(ind))
match = compute_overlap_and_matches(
path1, path2, self.max_dist_overlap)
if (self.verbose):
log.info(match['path_source'],
match['path_target'],
'overlap={}'.format(match['overlap']))
if (np.max(
match['overlap']) > self.min_overlap_ratio
and np.max(match['overlap']) <
self.max_overlap_ratio):
np.save(out_path, match)
ind += 1
def _compute_matches_between_fragments(self, mod):
out_dir = osp.join(self.processed_dir, mod, 'matches')
if files_exist([out_dir]): # pragma: no cover
return
makedirs(out_dir)
ind = 0
list_drive = self.dict_seq[mod]
for drive in list_drive:
path_fragment = osp.join(self.processed_dir, mod, "fragment",
"{:02d}".format(drive))
list_name_frames = sorted(
[f for f in os.listdir(path_fragment) if "pt" in f])
# pre_compute specific pair
log.info("Compute the pairs")
if (self.min_dist is not None):
pair_time_frame = compute_spaced_time_frame(
list_name_frames, path_fragment, self.min_dist)
else:
pair_time_frame = [
(i, j) for i in range(len(list_name_frames))
for j in range(len(list_name_frames))
if (j - i) > 0 and (j - i) < self.max_time_distance
]
log.info("Compute the matches")
for i, j in pair_time_frame:
out_path = osp.join(out_dir, 'matches{:06d}.npy'.format(ind))
path1 = osp.join(path_fragment, list_name_frames[i])
path2 = osp.join(path_fragment, list_name_frames[j])
data1 = torch.load(path1)
data2 = torch.load(path2)
match = compute_overlap_and_matches(data1, data2,
self.max_dist_overlap)
match['path_source'] = path1
match['path_target'] = path2
match['name_source'] = i
match['name_target'] = j
match['scene'] = drive
np.save(out_path, match)
ind += 1
