ScanCapture pose format

[Yzhbuaa]

Hello, I captured several datasets using the ScanCapture App. But I don't know the definition of the image pose in poses.txt file. Could you explain it in detail? Thank you!

[sarlinpe]

The pose is the transformation from camera frame to world frame but follows the computer graphics conventions: https://developer.apple.com/documentation/arkit/arcamera/2866108-transform
https://developer.apple.com/documentation/arkit/configuration_objects/understanding_world_tracking

the y-axis points upward, and (when relevant) the z-axis points toward the viewer and the x-axis points toward the viewer's right.

Here is how to parse it:

lamar-benchmark/scantools/run_phone_to_capture.py

Lines 106 to 118 in ede1c43

	# Poses from ARKit & ARCore use the computer graphics conventions (inverted y and z)
	rot_cg_to_cv = Rotation.from_matrix(np.diag([1, -1, -1]))
	rows = read_csv(path)
	for ts, status, tx, ty, tz, qx, qy, qz, qw, w, h, fx, fy, cx, cy, *_ in tqdm(rows):
	ts = int(ts)
	if status != 'normal':
	logger.warning('Tracking of %d with abnormal status: %s.', ts, status)
	tvec = np.array([tx, ty, tz], float)
	qvec = np.array([qw, qx, qy, qz], float)
	rot = Pose(qvec).r * rot_cg_to_cv
	pose = Pose(rot, tvec)