def export_openmvg_structure(
kapture_points_3d: kapture.Points3d,
kapture_to_openmvg_view_ids: Dict[str, int],
kapture_observations: Optional[kapture.Observations] = None,
kapture_keypoints: Optional[kapture.Keypoints] = None,
kapture_path: Optional[str] = None,
):
# early check
if kapture_points_3d is None:
logger.warning('no 3D points to export.')
return
xyz_coordinates = kapture_points_3d[:, 0:3]
include_2d_observations = kapture_observations is not None
openmvg_structure = []
# this loop can be very long, lets show some progress
hide_progress_bars = logger.getEffectiveLevel() > logging.INFO
for point_idx, coords in enumerate(tqdm(xyz_coordinates, disable=hide_progress_bars)):
point_3d_structure = {
'key': point_idx,
'value': {
'X': coords.tolist(),
'observations': []
}
}
if include_2d_observations and point_idx in kapture_observations:
for kapture_image_name, feature_point_id in kapture_observations[point_idx]:
openmvg_view_id = kapture_to_openmvg_view_ids[kapture_image_name]
point_2d_observation = {'key': openmvg_view_id,
'value': {'id_feat': feature_point_id, }}
if kapture_path and kapture_keypoints is not None:
# if given, load keypoints to populate 2D coordinates of the feature.
keypoints_file_path = get_keypoints_fullpath(kapture_path, kapture_image_name)
try:
keypoints_data = image_keypoints_from_file(keypoints_file_path,
dtype=kapture_keypoints.dtype,
dsize=kapture_keypoints.dsize)
point_2d_observation['value']['x'] = keypoints_data[feature_point_id, 0:2].tolist()
except FileNotFoundError:
logger.warning(f'unable to load keypoints file {keypoints_file_path}')
point_3d_structure['value']['observations'].append(point_2d_observation)
openmvg_structure.append(point_3d_structure)
return openmvg_structure
def extract_kapture_keypoints(kapture_root,
config,
output_dir='',
overwrite=False):
"""
Extract r2d2 keypoints and descritors to the kapture format directly
"""
print('extract_kapture_keypoints...')
kdata = kapture_from_dir(kapture_root, matches_pairsfile_path=None,
skip_list= [kapture.GlobalFeatures,
kapture.Matches,
kapture.Points3d,
kapture.Observations])
export_dir = output_dir if output_dir else kapture_root # root of output directory for features
os.makedirs(export_dir, exist_ok=True)
assert kdata.records_camera is not None
image_list = [filename for _, _, filename in kapture.flatten(kdata.records_camera)]
# resume extraction if some features exist
try:
# load existing features, if any
kdata.keypoints = keypoints_from_dir(export_dir, None)
kdata.descriptors = descriptors_from_dir(export_dir, None)
if kdata.keypoints is not None and kdata.descriptors is not None and not overwrite:
image_list = [name for name in image_list if name not in kdata.keypoints or name not in kdata.descriptors]
except FileNotFoundError:
pass
except:
logging.exception("Error with importing existing local features.")
# clear features first if overwriting
if overwrite: delete_existing_kapture_files(export_dir, True, only=[kapture.Descriptors, kapture.Keypoints])
if len(image_list) == 0:
print('All features were already extracted')
return
else:
print(f'Extracting r2d2 features for {len(image_list)} images')
iscuda = common.torch_set_gpu([torch.cuda.is_available()])
# load the network...
net = load_network(config['checkpoint'])
if iscuda: net = net.cuda()
# create the non-maxima detector
detector = NonMaxSuppression(
rel_thr = config['reliability_thr'],
rep_thr = config['repeatability_thr'])
keypoints_dtype = None if kdata.keypoints is None else kdata.keypoints.dtype
descriptors_dtype = None if kdata.descriptors is None else kdata.descriptors.dtype
keypoints_dsize = None if kdata.keypoints is None else kdata.keypoints.dsize
descriptors_dsize = None if kdata.descriptors is None else kdata.descriptors.dsize
for image_name in image_list:
img_path = get_image_fullpath(kapture_root, image_name)
if img_path.endswith('.txt'):
images = open(img_path).read().splitlines() + images
continue
print(f"\nExtracting features for {img_path}")
img = Image.open(img_path).convert('RGB')
W, H = img.size
img = norm_RGB(img)[None]
if iscuda: img = img.cuda()
# extract keypoints/descriptors for a single image
xys, desc, scores = extract_multiscale(net, img, detector,
scale_f = config['scale_f'],
min_scale = config['min_scale'],
max_scale = config['max_scale'],
min_size = config['min_size'],
max_size = config['max_size'],
verbose = True)
xys = xys.cpu().numpy()
desc = desc.cpu().numpy()
scores = scores.cpu().numpy()
idxs = scores.argsort()[-config['top_k'] or None:]
xys = xys[idxs]
desc = desc[idxs]
if keypoints_dtype is None or descriptors_dtype is None:
keypoints_dtype = xys.dtype
descriptors_dtype = desc.dtype
keypoints_dsize = xys.shape[1]
descriptors_dsize = desc.shape[1]
kdata.keypoints = kapture.Keypoints('r2d2', keypoints_dtype, keypoints_dsize)
kdata.descriptors = kapture.Descriptors('r2d2', descriptors_dtype, descriptors_dsize)
keypoints_config_absolute_path = get_csv_fullpath(kapture.Keypoints, export_dir)
descriptors_config_absolute_path = get_csv_fullpath(kapture.Descriptors, export_dir)
keypoints_to_file(keypoints_config_absolute_path, kdata.keypoints)
descriptors_to_file(descriptors_config_absolute_path, kdata.descriptors)
else:
assert kdata.keypoints.type_name == 'r2d2'
assert kdata.descriptors.type_name == 'r2d2'
assert kdata.keypoints.dtype == xys.dtype
assert kdata.descriptors.dtype == desc.dtype
assert kdata.keypoints.dsize == xys.shape[1]
assert kdata.descriptors.dsize == desc.shape[1]
keypoints_fullpath = get_keypoints_fullpath(export_dir, image_name)
print(f"Saving {xys.shape[0]} keypoints to {keypoints_fullpath}")
image_keypoints_to_file(keypoints_fullpath, xys)
kdata.keypoints.add(image_name)
descriptors_fullpath = get_descriptors_fullpath(export_dir, image_name)
print(f"Saving {desc.shape[0]} descriptors to {descriptors_fullpath}")
image_descriptors_to_file(descriptors_fullpath, desc)
kdata.descriptors.add(image_name)
if not keypoints_check_dir(kdata.keypoints, export_dir) or \
not descriptors_check_dir(kdata.descriptors, export_dir):
print('local feature extraction ended successfully but not all files were saved')
def export_opensfm(
kapture_rootdir: str,
opensfm_rootdir: str,
force_overwrite_existing: bool = False,
images_import_method: TransferAction = TransferAction.copy) -> None:
"""
:param kapture_rootdir:
:param opensfm_rootdir:
:param force_overwrite_existing:
:param images_import_method:
:return:
"""
disable_tqdm = logger.getEffectiveLevel(
) > logging.INFO # dont display tqdm for non-verbose levels
# load reconstruction
kapture_data = kapture.io.csv.kapture_from_dir(
kapture_dirpath=kapture_rootdir)
# export cameras
opensfm_cameras = {}
kapture_cameras = {
cam_id: cam
for cam_id, cam in kapture_data.sensors.items()
if cam.sensor_type == 'camera'
}
for cam_id, kapture_camera in kapture_cameras.items():
opensfm_cameras[cam_id] = export_opensfm_camera(kapture_camera)
# export shots
opensfm_shots = {}
for timestamp, camera_id, image_filename in tqdm(kapture.flatten(
kapture_data.records_camera),
disable=disable_tqdm):
# retrieve pose (if there is one).
# opensfm_shots = {image_filename: shot}
# shot = {camera , rotation, translation, capture_time, gps_position, ...}
opensfm_shot = {
'capture_time': 0, # in ms != timestamp
'camera': camera_id,
}
if (timestamp, camera_id) in kapture_data.trajectories:
pose = kapture_data.trajectories[timestamp, camera_id]
rotation_vector = quaternion.as_rotation_vector(pose.r)
translation_vector = pose.t.flatten()
opensfm_shot.update({
'rotation': rotation_vector.tolist(),
'translation': translation_vector.tolist()
})
opensfm_shots[image_filename] = opensfm_shot
# pack it
opensfm_reconstruction = {
'cameras': opensfm_cameras,
'shots': opensfm_shots,
}
# images
logger.info(
f'writing image files "{path.join(opensfm_rootdir, "images")}".')
image_filenames = [
f for _, _, f in kapture.flatten(kapture_data.records_camera)
]
kapture_image_filepaths = [
get_record_fullpath(kapture_rootdir, image_filename)
for image_filename in image_filenames
]
opensfm_image_filepaths = [
path.join(opensfm_rootdir, 'images', image_filename)
for image_filename in image_filenames
]
transfer_files_from_dir(
source_filepath_list=kapture_image_filepaths,
destination_filepath_list=opensfm_image_filepaths,
force_overwrite=force_overwrite_existing,
copy_strategy=images_import_method,
)
# export features files (keypoints + descriptors)
opensfm_features_suffix = '.features.npz'
opensfm_features_dirpath = path.join(opensfm_rootdir, 'features')
logger.info(
f'exporting keypoint and descriptors to {opensfm_features_dirpath}')
os.makedirs(opensfm_features_dirpath, exist_ok=True)
for image_filename in tqdm(image_filenames, disable=disable_tqdm):
opensfm_features = {}
# look and load for keypoints in kapture
if kapture_data.keypoints is not None and image_filename in kapture_data.keypoints:
kapture_keypoints_filepath = get_keypoints_fullpath(
kapture_dirpath=kapture_rootdir, image_filename=image_filename)
logger.debug(f'loading {kapture_keypoints_filepath}')
kapture_keypoint = image_keypoints_from_file(
kapture_keypoints_filepath,
dtype=kapture_data.keypoints.dtype,
dsize=kapture_data.keypoints.dsize)
opensfm_features['points'] = kapture_keypoint
# look and load for descriptors in kapture
if kapture_data.descriptors is not None and image_filename in kapture_data.descriptors:
kapture_descriptor_filepath = get_descriptors_fullpath(
kapture_dirpath=kapture_rootdir, image_filename=image_filename)
logger.debug(f'loading {kapture_descriptor_filepath}')
kapture_descriptor = image_descriptors_from_file(
kapture_descriptor_filepath,
dtype=kapture_data.descriptors.dtype,
dsize=kapture_data.descriptors.dsize)
opensfm_features['descriptors'] = kapture_descriptor
# writing opensfm feature file
if len(opensfm_features) > 0:
opensfm_features_filepath = path.join(
opensfm_features_dirpath,
image_filename + opensfm_features_suffix)
logger.debug(f'writing {opensfm_features_filepath}')
os.makedirs(path.dirname(opensfm_features_filepath), exist_ok=True)
np.save(opensfm_features_filepath, opensfm_features)
# export matches files
if kapture_data.matches is not None:
opensfm_matches_suffix = '_matches.pkl.gz'
opensfm_matches_dirpath = path.join(opensfm_rootdir, 'matches')
os.makedirs(opensfm_matches_dirpath, exist_ok=True)
logger.info(f'exporting matches to {opensfm_matches_dirpath}')
opensfm_pairs = {}
for image_filename1, image_filename2 in kapture_data.matches:
opensfm_pairs.setdefault(image_filename1,
[]).append(image_filename2)
for image_filename1 in tqdm(image_filenames, disable=disable_tqdm):
opensfm_matches = {}
opensfm_matches_filepath = path.join(
opensfm_matches_dirpath,
image_filename1 + opensfm_matches_suffix)
logger.debug(f'loading matches for {image_filename1}')
for image_filename2 in opensfm_pairs.get(image_filename1, []):
# print(image_filename1, image_filename2)
kapture_matches_filepath = get_matches_fullpath(
(image_filename1, image_filename2),
kapture_dirpath=kapture_rootdir)
kapture_matches = image_matches_from_file(
kapture_matches_filepath)
opensfm_matches[image_filename2] = kapture_matches[:,
0:2].astype(
np.int)
os.makedirs(path.dirname(opensfm_matches_filepath), exist_ok=True)
with gzip.open(opensfm_matches_filepath, 'wb') as f:
pickle.dump(opensfm_matches, f)
# export 3D-points files
if kapture_data.points3d is not None:
logger.info('exporting points 3-D')
opensfm_reconstruction['points'] = {}
for i, (x, y, z, r, g, b) in tqdm(enumerate(kapture_data.points3d),
disable=disable_tqdm):
opensfm_reconstruction['points'][i] = {
'coordinates': [x, y, z],
'color': [r, g, b]
}
# write json files #################################################################################################
os.makedirs(opensfm_rootdir, exist_ok=True)
# write reconstruction.json
opensfm_reconstruction_filepath = path.join(opensfm_rootdir,
'reconstruction.json')
logger.info(
f'writing reconstruction file "{opensfm_reconstruction_filepath}".')
with open(opensfm_reconstruction_filepath, 'wt') as f:
json.dump([opensfm_reconstruction], f, indent=4)
# write camera_models.json
opensfm_cameras_filepath = path.join(opensfm_rootdir, 'camera_models.json')
logger.info(f'writing camera models file "{opensfm_cameras_filepath}".')
with open(opensfm_cameras_filepath, 'wt') as f:
json.dump(opensfm_cameras, f, indent=4)
kdata.keypoints = kapture.Keypoints('d2net', keypoints_dtype, keypoints_dsize)
kdata.descriptors = kapture.Descriptors('d2net', descriptors_dtype, descriptors_dsize)
keypoints_config_absolute_path = get_csv_fullpath(kapture.Keypoints, args.kapture_root)
descriptors_config_absolute_path = get_csv_fullpath(kapture.Descriptors, args.kapture_root)
keypoints_to_file(keypoints_config_absolute_path, kdata.keypoints)
descriptors_to_file(descriptors_config_absolute_path, kdata.descriptors)
else:
assert kdata.keypoints.type_name == 'd2net'
assert kdata.descriptors.type_name == 'd2net'
assert kdata.keypoints.dtype == keypoints.dtype
assert kdata.descriptors.dtype == descriptors.dtype
assert kdata.keypoints.dsize == keypoints.shape[1]
assert kdata.descriptors.dsize == descriptors.shape[1]
keypoints_fullpath = get_keypoints_fullpath(args.kapture_root, image_name)
print(f"Saving {keypoints.shape[0]} keypoints to {keypoints_fullpath}")
image_keypoints_to_file(keypoints_fullpath, keypoints)
kdata.keypoints.add(image_name)
descriptors_fullpath = get_descriptors_fullpath(args.kapture_root, image_name)
print(f"Saving {descriptors.shape[0]} descriptors to {descriptors_fullpath}")
image_descriptors_to_file(descriptors_fullpath, descriptors)
kdata.descriptors.add(image_name)
if not keypoints_check_dir(kdata.keypoints, args.kapture_root) or \
not descriptors_check_dir(kdata.descriptors, args.kapture_root):
print('local feature extraction ended successfully but not all files were saved')
def add_frames(self, frames: List[Frame], points3d: List[Keypoint]):
k = self.kapture
if k.records_camera is None:
k.records_camera = kt.RecordsCamera()
if k.trajectories is None:
k.trajectories = kt.Trajectories()
if k.keypoints is None:
k.keypoints = {
self.default_kp_type:
kt.Keypoints(self.default_kp_type, np.float32, 2)
}
if k.points3d is None:
k.points3d = kt.Points3d()
if k.observations is None:
k.observations = kt.Observations()
def check_kp(kp):
return not kp.bad_qlt and kp.inlier_count > self.min_pt3d_obs and kp.inlier_count / kp.total_count > self.min_pt3d_ratio
kp_ids, pts3d = zip(*[(kp.id, kp.pt3d) for kp in points3d
if check_kp(kp)])
I = np.argsort(kp_ids)
pt3d_ids = dict(zip(np.array(kp_ids)[I], np.arange(len(I))))
pt3d_arr = np.array(pts3d)[I, :]
k.points3d = kt.Points3d(
np.concatenate((pt3d_arr, np.ones_like(pt3d_arr) * 128), axis=1))
for f in frames:
if not f.pose.post:
continue
id = f.frame_num
img = f.orig_image
img_file = os.path.join(self.default_cam[1],
'frame%06d.%s' % (id, self.img_format))
img_fullpath = get_record_fullpath(self.path, img_file)
os.makedirs(os.path.dirname(img_fullpath), exist_ok=True)
if not np.isclose(self.scale, 1.0):
img = cv2.resize(img,
None,
fx=self.scale,
fy=self.scale,
interpolation=cv2.INTER_AREA)
if self.img_format == self.IMG_FORMAT_PNG:
cv2.imwrite(img_fullpath, img,
(cv2.IMWRITE_PNG_COMPRESSION, 9))
elif self.img_format == self.IMG_FORMAT_JPG:
cv2.imwrite(img_fullpath, img,
(cv2.IMWRITE_JPEG_QUALITY, self.jpg_qlt))
else:
assert False, 'Invalid image format: %s' % (self.img_format, )
record_id = (id, self.default_cam[0])
k.records_camera[record_id] = img_file
pose = f.pose.post if 1 else (-f.pose.post)
k.trajectories[record_id] = kt.PoseTransform(
r=pose.quat.components, t=pose.loc)
k.keypoints[self.default_kp_type].add(img_file)
uvs = np.zeros((len(f.kps_uv), 2), np.float32)
i = 0
for kp_id, uv in f.kps_uv.items():
if kp_id in pt3d_ids:
k.observations.add(int(pt3d_ids[kp_id]),
self.default_kp_type, img_file, i)
uvs[i, :] = uv / f.img_sc * self.scale
i += 1
image_keypoints_to_file(
get_keypoints_fullpath(self.default_kp_type, self.path,
img_file), uvs[:i, :])
def extract_kapture_keypoints(args):
"""
Extract r2d2 keypoints and descritors to the kapture format directly
"""
print('extract_kapture_keypoints...')
kdata = kapture_from_dir(args.kapture_root,
matches_pairsfile_path=None,
skip_list=[
kapture.GlobalFeatures, kapture.Matches,
kapture.Points3d, kapture.Observations
])
assert kdata.records_camera is not None
image_list = [
filename for _, _, filename in kapture.flatten(kdata.records_camera)
]
if kdata.keypoints is not None and kdata.descriptors is not None:
image_list = [
name for name in image_list
if name not in kdata.keypoints or name not in kdata.descriptors
]
if len(image_list) == 0:
print('All features were already extracted')
return
else:
print(f'Extracting r2d2 features for {len(image_list)} images')
iscuda = common.torch_set_gpu(args.gpu)
# load the network...
net = load_network(args.model)
if iscuda: net = net.cuda()
# create the non-maxima detector
detector = NonMaxSuppression(rel_thr=args.reliability_thr,
rep_thr=args.repeatability_thr)
keypoints_dtype = None if kdata.keypoints is None else kdata.keypoints.dtype
descriptors_dtype = None if kdata.descriptors is None else kdata.descriptors.dtype
keypoints_dsize = None if kdata.keypoints is None else kdata.keypoints.dsize
descriptors_dsize = None if kdata.descriptors is None else kdata.descriptors.dsize
for image_name in image_list:
img_path = get_image_fullpath(args.kapture_root, image_name)
print(f"\nExtracting features for {img_path}")
img = Image.open(img_path).convert('RGB')
W, H = img.size
img = norm_RGB(img)[None]
if iscuda: img = img.cuda()
# extract keypoints/descriptors for a single image
xys, desc, scores = extract_multiscale(net,
img,
detector,
scale_f=args.scale_f,
min_scale=args.min_scale,
max_scale=args.max_scale,
min_size=args.min_size,
max_size=args.max_size,
verbose=True)
xys = xys.cpu().numpy()
desc = desc.cpu().numpy()
scores = scores.cpu().numpy()
idxs = scores.argsort()[-args.top_k or None:]
xys = xys[idxs]
desc = desc[idxs]
if keypoints_dtype is None or descriptors_dtype is None:
keypoints_dtype = xys.dtype
descriptors_dtype = desc.dtype
keypoints_dsize = xys.shape[1]
descriptors_dsize = desc.shape[1]
kdata.keypoints = kapture.Keypoints('r2d2', keypoints_dtype,
keypoints_dsize)
kdata.descriptors = kapture.Descriptors('r2d2', descriptors_dtype,
descriptors_dsize)
keypoints_config_absolute_path = get_csv_fullpath(
kapture.Keypoints, args.kapture_root)
descriptors_config_absolute_path = get_csv_fullpath(
kapture.Descriptors, args.kapture_root)
keypoints_to_file(keypoints_config_absolute_path, kdata.keypoints)
descriptors_to_file(descriptors_config_absolute_path,
kdata.descriptors)
else:
assert kdata.keypoints.type_name == 'r2d2'
assert kdata.descriptors.type_name == 'r2d2'
assert kdata.keypoints.dtype == xys.dtype
assert kdata.descriptors.dtype == desc.dtype
assert kdata.keypoints.dsize == xys.shape[1]
assert kdata.descriptors.dsize == desc.shape[1]
keypoints_fullpath = get_keypoints_fullpath(args.kapture_root,
image_name)
print(f"Saving {xys.shape[0]} keypoints to {keypoints_fullpath}")
image_keypoints_to_file(keypoints_fullpath, xys)
kdata.keypoints.add(image_name)
descriptors_fullpath = get_descriptors_fullpath(
args.kapture_root, image_name)
print(f"Saving {desc.shape[0]} descriptors to {descriptors_fullpath}")
image_descriptors_to_file(descriptors_fullpath, desc)
kdata.descriptors.add(image_name)
if not keypoints_check_dir(kdata.keypoints, args.kapture_root) or \
not descriptors_check_dir(kdata.descriptors, args.kapture_root):
print(
'local feature extraction ended successfully but not all files were saved'
)
def load_keypoints(keypoints_type, input_path, image_name, dtype, dsize,
tar_handlers):
keypoints_path = get_keypoints_fullpath(keypoints_type, input_path,
image_name, tar_handlers)
return image_keypoints_from_file(keypoints_path, dtype, dsize)
kdata.descriptors[args.descriptors_type])
else:
assert kdata.keypoints[
args.keypoints_type].dtype == keypoints.dtype
assert kdata.descriptors[
args.descriptors_type].dtype == descriptors.dtype
assert kdata.keypoints[
args.keypoints_type].dsize == keypoints.shape[1]
assert kdata.descriptors[
args.descriptors_type].dsize == descriptors.shape[1]
assert kdata.descriptors[
args.descriptors_type].keypoints_type == args.keypoints_type
assert kdata.descriptors[args.descriptors_type].metric_type == 'L2'
keypoints_fullpath = get_keypoints_fullpath(args.keypoints_type,
args.kapture_root,
image_name, tar_handlers)
print(f"Saving {keypoints.shape[0]} keypoints to {keypoints_fullpath}")
image_keypoints_to_file(keypoints_fullpath, keypoints)
kdata.keypoints[args.keypoints_type].add(image_name)
descriptors_fullpath = get_descriptors_fullpath(
args.descriptors_type, args.kapture_root, image_name, tar_handlers)
print(
f"Saving {descriptors.shape[0]} descriptors to {descriptors_fullpath}"
)
image_descriptors_to_file(descriptors_fullpath, descriptors)
kdata.descriptors[args.descriptors_type].add(image_name)
if not keypoints_check_dir(kdata.keypoints[args.keypoints_type], args.keypoints_type,
args.kapture_root, tar_handlers) or \
def _export_opensfm_features_and_matches(image_filenames, kapture_data,
kapture_root_dir, opensfm_root_dir,
disable_tqdm):
"""
export features files (keypoints + descriptors) and matches
"""
opensfm_features_suffix = '.features.npz'
opensfm_features_dir_path = path.join(opensfm_root_dir, 'features')
logger.info(
f'exporting keypoint and descriptors to {opensfm_features_dir_path}')
os.makedirs(opensfm_features_dir_path, exist_ok=True)
for image_filename in tqdm(image_filenames, disable=disable_tqdm):
opensfm_features = {}
# look and load for keypoints in kapture
if kapture_data.keypoints is not None and image_filename in kapture_data.keypoints:
kapture_keypoints_filepath = get_keypoints_fullpath(
kapture_dirpath=kapture_root_dir,
image_filename=image_filename)
logger.debug(f'loading {kapture_keypoints_filepath}')
kapture_keypoint = image_keypoints_from_file(
kapture_keypoints_filepath,
dtype=kapture_data.keypoints.dtype,
dsize=kapture_data.keypoints.dsize)
opensfm_features['points'] = kapture_keypoint
# look and load for descriptors in kapture
if kapture_data.descriptors is not None and image_filename in kapture_data.descriptors:
kapture_descriptor_filepath = get_descriptors_fullpath(
kapture_dirpath=kapture_root_dir,
image_filename=image_filename)
logger.debug(f'loading {kapture_descriptor_filepath}')
kapture_descriptor = image_descriptors_from_file(
kapture_descriptor_filepath,
dtype=kapture_data.descriptors.dtype,
dsize=kapture_data.descriptors.dsize)
opensfm_features['descriptors'] = kapture_descriptor
# writing opensfm feature file
if len(opensfm_features) > 0:
opensfm_features_filepath = path.join(
opensfm_features_dir_path,
image_filename + opensfm_features_suffix)
logger.debug(f'writing {opensfm_features_filepath}')
os.makedirs(path.dirname(opensfm_features_filepath), exist_ok=True)
np.save(opensfm_features_filepath, opensfm_features)
# export matches files
if kapture_data.matches is not None:
opensfm_matches_suffix = '_matches.pkl.gz'
opensfm_matches_dir_path = path.join(opensfm_root_dir, 'matches')
os.makedirs(opensfm_matches_dir_path, exist_ok=True)
logger.info(f'exporting matches to {opensfm_matches_dir_path}')
opensfm_pairs = {}
for image_filename1, image_filename2 in kapture_data.matches:
opensfm_pairs.setdefault(image_filename1,
[]).append(image_filename2)
for image_filename1 in tqdm(image_filenames, disable=disable_tqdm):
opensfm_matches = {}
opensfm_matches_filepath = path.join(
opensfm_matches_dir_path,
image_filename1 + opensfm_matches_suffix)
logger.debug(f'loading matches for {image_filename1}')
for image_filename2 in opensfm_pairs.get(image_filename1, []):
# print(image_filename1, image_filename2)
kapture_matches_filepath = get_matches_fullpath(
(image_filename1, image_filename2),
kapture_dirpath=kapture_root_dir)
kapture_matches = image_matches_from_file(
kapture_matches_filepath)
opensfm_matches[image_filename2] = kapture_matches[:,
0:2].astype(
np.int)
os.makedirs(path.dirname(opensfm_matches_filepath), exist_ok=True)
with gzip.open(opensfm_matches_filepath, 'wb') as f:
pickle.dump(opensfm_matches, f)
def load_from_kapture(self, kapture_data, minimal_observation_count=10):
image_list = [
filename
for _, _, filename in kapture.flatten(kapture_data.records_camera)
]
descriptors = []
keypoints = []
points3d = []
mask = []
image_indexes = {}
image_index_list = []
keypoint_index_list = []
self.keypoint_count = 0
for i, image_path in enumerate(image_list):
descriptors_full_path = get_descriptors_fullpath(
self._descriptor_name, kapture_data.kapture_path, image_path)
kapture_descriptors = kapture_data.descriptors[
self._descriptor_name]
descriptors.append(
image_descriptors_from_file(descriptors_full_path,
kapture_descriptors.dtype,
kapture_descriptors.dsize))
keypoints_full_path = get_keypoints_fullpath(
self._descriptor_name, kapture_data.kapture_path, image_path)
kapture_keypoints = kapture_data.keypoints[self._descriptor_name]
keypoints.append(
image_keypoints_from_file(keypoints_full_path,
kapture_keypoints.dtype,
kapture_keypoints.dsize))
point_count = len(keypoints[i])
points3d.append(np.zeros((point_count, 3), dtype=np.float32))
mask.append(np.zeros(point_count, dtype=np.bool))
image_indexes[image_path] = i
image_index_list.append(np.array([i] * point_count))
keypoint_index_list.append(np.arange(point_count))
self.keypoint_count += point_count
for point_index, observation in kapture_data.observations.items():
if len(observation[
self._descriptor_name]) > minimal_observation_count:
for observation_image_name, image_keypoints_index in observation[
self._descriptor_name]:
image_index = image_indexes[observation_image_name]
mask[image_index][image_keypoints_index] = True
points3d[image_index][
image_keypoints_index] = kapture_data.points3d[
point_index][:3]
for i in range(len(mask)):
self.image_index_list.extend(list(image_index_list[i]))
self.keypoint_index_list.extend(list(keypoint_index_list[i]))
self.masked_image_index_list.extend(
list(image_index_list[i][mask[i]]))
self.masked_keypoint_index_list.extend(
list(keypoint_index_list[i][mask[i]]))
self.descriptors = descriptors
self.keypoints = keypoints
self.points3d = points3d
self.mask = mask
self.image_index_from_image_name = image_indexes
self.load_trajectory(kapture_data)