def convert_cameras(id_to_col_cameras, id_to_col_images, image_dp,
image_fp_type, suppress_distortion_warnings, op):
# From photogrammetry_importer\ext\read_write_model.py
# CameraModel = collections.namedtuple(
# "CameraModel", ["model_id", "model_name", "num_params"])
# Camera = collections.namedtuple(
# "Camera", ["id", "model", "width", "height", "params"])
# BaseImage = collections.namedtuple(
# "Image", ["id", "qvec", "tvec", "camera_id", "name", "xys", "point3D_ids"])
cameras = []
for col_image in id_to_col_images.values():
current_camera = Camera()
current_camera.id = col_image.id
current_camera.set_quaternion(col_image.qvec)
current_camera.set_camera_translation_vector_after_rotation(
col_image.tvec)
current_camera.image_fp_type = image_fp_type
current_camera.image_dp = image_dp
current_camera._relative_fp = col_image.name
camera_model = id_to_col_cameras[col_image.camera_id]
# op.report({'INFO'}, 'image_id: ' + str(col_image.id))
# op.report({'INFO'}, 'camera_id: ' + str(col_image.camera_id))
# op.report({'INFO'}, 'camera_model: ' + str(camera_model))
current_camera.width = camera_model.width
current_camera.height = camera_model.height
focal_length, cx, cy, r = parse_camera_param_list(camera_model)
if not suppress_distortion_warnings:
check_radial_distortion(r, current_camera._relative_fp, op)
camera_calibration_matrix = np.array([[focal_length, 0, cx],
[0, focal_length, cy],
[0, 0, 1]])
current_camera.set_calibration(camera_calibration_matrix,
radial_distortion=0)
cameras.append(current_camera)
return cameras
def parse_cameras(json_data, image_dp, image_fp_type,
suppress_distortion_warnings, op):
json_cameras_intrinsics = json_data['cameras']
views = json_data['shots']
cams = []
for view_name in views:
view = views[view_name]
camera = Camera()
camera.image_fp_type = image_fp_type
camera.image_dp = image_dp
camera._relative_fp = view_name
camera._absolute_fp = os.path.join(image_dp, view_name)
intrinsic_key = view['camera']
focal_length, cx, cy, width, height, radial_distortion = OpenSfMJSONFileHandler.convert_intrinsics(
json_cameras_intrinsics[intrinsic_key], camera._relative_fp,
suppress_distortion_warnings, op)
camera.height = height
camera.width = width
camera_calibration_matrix = np.array([[focal_length, 0, cx],
[0, focal_length, cy],
[0, 0, 1]])
camera.set_calibration(camera_calibration_matrix,
radial_distortion)
rodrigues_vec = np.array(view['rotation'], dtype=float)
rot_mat = OpenSfMJSONFileHandler.rodrigues_to_matrix(rodrigues_vec)
camera.set_rotation_mat(rot_mat)
camera.set_camera_translation_vector_after_rotation(
np.array(view['translation'], dtype=float))
cams.append(camera)
return cams
def _parse_open3d_log_file(open3d_ifp, image_dp, image_relative_fp_list, image_fp_type, op):
cams = []
with open(open3d_ifp, 'r') as open3d_file:
lines = open3d_file.readlines()
# Chunk size: 1 line meta data, 4 lines for the matrix
chunk_size = 5
assert len(lines) % chunk_size == 0
chunks = Open3DFileHandler._chunker(lines, chunk_size)
if len(chunks) != len(image_relative_fp_list):
# Create some dummy names for missing images
image_relative_fp_list = Open3DFileHandler._create_dummy_fp_list(
len(chunks))
for chunk, image_relative_fp in zip(chunks, image_relative_fp_list):
meta_data = chunk[0]
matrix_list = [
Open3DFileHandler._read_matrix_row(chunk[1]),
Open3DFileHandler._read_matrix_row(chunk[2]),
Open3DFileHandler._read_matrix_row(chunk[3])]
# Note: the transformation matrix in the .json file is the inverse of
# the transformation matrix in the .log file
extrinsic_matrix = np.asarray(matrix_list, dtype=float)
cam = Camera()
cam.image_fp_type = image_fp_type
cam.image_dp = image_dp
cam._relative_fp = image_relative_fp
image_absolute_fp = os.path.join(image_dp, image_relative_fp)
cam._absolute_fp = image_absolute_fp
# Accuracy of rotation matrices is too low => disable rotation test
cam.set_4x4_cam_to_world_mat(
extrinsic_matrix, check_rotation=False)
cams.append(cam)
return cams
def _parse_open3d_json_file(open3d_ifp, image_dp, image_relative_fp_list, image_fp_type, op):
cams = []
with open(open3d_ifp, 'r') as open3d_file:
json_data = json.load(open3d_file)
parameters = json_data['parameters']
if len(parameters) != len(image_relative_fp_list):
# Create some dummy names for missing images
image_relative_fp_list = Open3DFileHandler._create_dummy_fp_list(
len(parameters))
for pinhole_camera_parameter, image_relative_fp in zip(parameters, image_relative_fp_list):
cam = Camera()
cam.image_fp_type = image_fp_type
cam.image_dp = image_dp
cam._relative_fp = image_relative_fp
cam._absolute_fp = os.path.join(image_dp, image_relative_fp)
extrinsic = pinhole_camera_parameter['extrinsic']
# Note: the transformation matrix in the .json file is the inverse of
# the transformation matrix in the .log file
extrinsic_mat = np.linalg.inv(
np.array(extrinsic, dtype=float).reshape((4,4)).T)
intrinsic = pinhole_camera_parameter['intrinsic']
cam.width = intrinsic['width']
cam.height = intrinsic['height']
# Accuracy of rotation matrices is too low => disable test
cam.set_4x4_cam_to_world_mat(extrinsic_mat, check_rotation=False)
intrinsic = intrinsic['intrinsic_matrix']
intrinsic_mat = np.array(intrinsic, dtype=float).reshape((3,3)).T
cam.set_calibration_mat(intrinsic_mat)
cams.append(cam)
return cams
def _parse_cameras_from_json_data(json_data, image_dp, image_fp_type,
suppress_distortion_warnings, op):
cams = []
image_index_to_camera_index = {}
is_valid_file = 'views' in json_data and 'intrinsics' in json_data and 'poses' in json_data
if not is_valid_file:
op.report({
'ERROR'
}, 'FILE FORMAT ERROR: Incorrect SfM/JSON file. Must contain the SfM reconstruction results: '
+ 'view, intrinsics and poses.')
return cams, image_index_to_camera_index
views = json_data['views'] # is a list of dicts (view)
intrinsics = json_data['intrinsics'] # is a list of dicts (intrinsic)
extrinsics = json_data['poses'] # is a list of dicts (extrinsic)
# IMPORTANT:
# Views contain the number of input images
# Extrinsics may contain only a subset of views!
# (Not all views are necessarily contained in the reconstruction)
for rec_index, extrinsic in enumerate(extrinsics):
camera = Camera()
view_index = int(extrinsic['poseId'])
image_index_to_camera_index[view_index] = rec_index
corresponding_view = get_element(views, "poseId", view_index, op)
camera.image_fp_type = image_fp_type
camera.image_dp = image_dp
camera._absolute_fp = str(corresponding_view['path'])
camera._relative_fp = os.path.basename(
str(corresponding_view['path']))
camera._undistorted_relative_fp = str(extrinsic['poseId']) + '.exr'
if image_dp is None:
camera._undistorted_absolute_fp = None
else:
camera._undistorted_absolute_fp = os.path.join(
image_dp, camera._undistorted_relative_fp)
camera.width = int(corresponding_view['width'])
camera.height = int(corresponding_view['height'])
id_intrinsic = int(corresponding_view['intrinsicId'])
intrinsic_params = get_element(intrinsics, "intrinsicId",
id_intrinsic, op)
focal_length = float(intrinsic_params['pxFocalLength'])
cx = float(intrinsic_params['principalPoint'][0])
cy = float(intrinsic_params['principalPoint'][1])
if 'distortionParams' in intrinsic_params and len(
intrinsic_params['distortionParams']) > 0:
# TODO proper handling of distortion parameters
radial_distortion = float(
intrinsic_params['distortionParams'][0])
else:
radial_distortion = 0.0
if not suppress_distortion_warnings:
check_radial_distortion(radial_distortion, camera._relative_fp,
op)
camera_calibration_matrix = np.array([[focal_length, 0, cx],
[0, focal_length, cy],
[0, 0, 1]])
camera.set_calibration(camera_calibration_matrix,
radial_distortion)
extrinsic_params = extrinsic['pose']['transform']
cam_rotation_list = extrinsic_params['rotation']
camera.set_rotation_mat(
np.array(cam_rotation_list, dtype=float).reshape(3, 3).T)
camera.set_camera_center_after_rotation(
np.array(extrinsic_params['center'], dtype=float))
camera.view_index = view_index
cams.append(camera)
return cams, image_index_to_camera_index
def parse_cameras(json_data, image_dp, image_fp_type,
suppress_distortion_warnings, op):
views = {item['key']: item for item in json_data['views']}
intrinsics = {item['key']: item for item in json_data['intrinsics']}
extrinsics = {item['key']: item for item in json_data['extrinsics']}
# IMPORTANT:
# Views contain the description about the dataset and attribute to Pose and Intrinsic data.
# View -> id_pose, id_intrinsic
# Since sometimes some views cannot be localized, there is some missing pose and intrinsic data.
# Extrinsics may contain only a subset of views! (Potentially not all views are contained in the reconstruction)
cams = []
# Iterate over views, and create camera if Intrinsic and Pose data exist
for id, view in views.items(): # Iterate over views
id_view = view[
'key'] # Should be equal to view['value']['ptr_wrapper']['data']['id_view']
view_data = view['value']['ptr_wrapper']['data']
id_pose = view_data['id_pose']
id_intrinsic = view_data['id_intrinsic']
# Check if the view is having corresponding Pose and Intrinsic data
if id_pose in extrinsics.keys() and \
id_intrinsic in intrinsics.keys():
camera = Camera()
camera.image_fp_type = image_fp_type
camera.image_dp = image_dp
camera._relative_fp = os.path.join(view_data['local_path'],
view_data['filename'])
camera._absolute_fp = os.path.join(json_data['root_path'],
view_data['local_path'],
view_data['filename'])
camera.width = view_data['width']
camera.height = view_data['height']
id_intrinsic = view_data['id_intrinsic']
# handle intrinsic params
intrinsic_data = intrinsics[int(
id_intrinsic)]['value']['ptr_wrapper']['data']
polymorphic_name = intrinsics[int(
id_intrinsic)]['value']['polymorphic_name']
if polymorphic_name == 'spherical':
camera.set_panoramic_type(
Camera.panoramic_type_equirectangular)
# create some dummy values
focal_length = 0
cx = camera.width / 2
cy = camera.height / 2
else:
focal_length = intrinsic_data['focal_length']
principal_point = intrinsic_data['principal_point']
cx = principal_point[0]
cy = principal_point[1]
# For Radial there are several options: "None", disto_k1, disto_k3
if 'disto_k3' in intrinsic_data:
radial_distortion = [
float(intrinsic_data['disto_k3'][0]),
float(intrinsic_data['disto_k3'][1]),
float(intrinsic_data['disto_k3'][2])
]
elif 'disto_k1' in intrinsic_data:
radial_distortion = float(intrinsic_data['disto_k1'][0])
else: # No radial distortion, i.e. pinhole camera model
radial_distortion = 0
if not suppress_distortion_warnings:
check_radial_distortion(radial_distortion,
camera._relative_fp, op)
camera_calibration_matrix = np.array([[focal_length, 0, cx],
[0, focal_length, cy],
[0, 0, 1]])
camera.set_calibration(camera_calibration_matrix,
radial_distortion)
extrinsic_params = extrinsics[id_pose]
cam_rotation_list = extrinsic_params['value']['rotation']
camera.set_rotation_mat(
np.array(cam_rotation_list, dtype=float))
camera.set_camera_center_after_rotation(
np.array(extrinsic_params['value']['center'], dtype=float))
camera.view_index = id_view
cams.append(camera)
return cams
def _parse_cameras(input_file, num_cameras, camera_calibration_matrix,
image_dp, image_fp_type, op):
"""
VisualSFM CAMERA coordinate system is the standard CAMERA coordinate system in computer vision (not the same
as in computer graphics like in bundler, blender, etc.)
That means
the y axis in the image is pointing downwards (not upwards)
the camera is looking along the positive z axis (points in front of the camera show a positive z value)
The camera coordinate system in computer vision VISUALSFM uses camera matrices,
which are rotated around the x axis by 180 degree
i.e. the y and z axis of the CAMERA MATRICES are inverted
therefore, the y and z axis of the TRANSLATION VECTOR are also inverted
"""
# op.report({'INFO'}, '_parse_cameras: ...')
cameras = []
for i in range(num_cameras):
line = input_file.readline()
# Read the camera section
# From the docs:
# <Camera> = <File name> <focal length> <quaternion WXYZ> <camera center> <radial distortion> 0
line_values = line.split()
relative_path = line_values[0].replace('/', os.sep)
focal_length = float(line_values[1])
quaternion_w = float(line_values[2])
quaternion_x = float(line_values[3])
quaternion_y = float(line_values[4])
quaternion_z = float(line_values[5])
quaternion = np.array(
[quaternion_w, quaternion_x, quaternion_y, quaternion_z],
dtype=float)
camera_center_x = float(line_values[6])
camera_center_y = float(line_values[7])
camera_center_z = float(line_values[8])
center_vec = np.array(
[camera_center_x, camera_center_y, camera_center_z])
radial_distortion = float(line_values[9])
if camera_calibration_matrix is None:
# In this case, we have no information about the principal point
# We assume that the principal point lies in the center
camera_calibration_matrix = np.array([[focal_length, 0, 0],
[0, focal_length, 0],
[0, 0, 1]])
zero_value = float(line_values[10])
assert (zero_value == 0)
current_camera = Camera()
# Setting the quaternion also sets the rotation matrix
current_camera.set_quaternion(quaternion)
# Set the camera center after rotation
# COMMENT FROM PBA CODE:
# older format for compability
# camera_data[i].SetQuaternionRotation(q); // quaternion from the file
# camera_data[i].SetCameraCenterAfterRotation(c); // camera center from the file
current_camera._center = center_vec
# set the camera view direction as normal w.r.t world coordinates
cam_view_vec_cam_coord = np.array([0, 0, 1]).T
cam_rotation_matrix_inv = np.linalg.inv(
current_camera.get_rotation_mat())
cam_view_vec_world_coord = cam_rotation_matrix_inv.dot(
cam_view_vec_cam_coord)
current_camera.normal = cam_view_vec_world_coord
translation_vec = NVMFileHandler.compute_camera_coordinate_system_translation_vector(
center_vec, current_camera.get_rotation_mat())
current_camera._translation_vec = translation_vec
current_camera.set_calibration(camera_calibration_matrix,
radial_distortion=radial_distortion)
# op.report({'INFO'}, 'Calibration mat:')
# op.report({'INFO'}, str(camera_calibration_matrix))
current_camera.image_fp_type = image_fp_type
current_camera.image_dp = image_dp
current_camera._relative_fp = relative_path
current_camera.id = i
cameras.append(current_camera)
# op.report({'INFO'}, '_parse_cameras: Done')
return cameras