L

lss5fc0d

If I use YoloSpatialDetectionNetwork to squeeze or letterbox my image (as described here) and then feed it to
YoloSpatialDetectionNetwork, should I expect the depth to work? It seems like there's no way for it to know how to align the disparity map. I've tried to figure this out experimentally but the results have been ambiguous. Is there a correct way to get full FOV and get the correct depth out of a Detection Network?

For orientation, I've had some success turning the ROTATION_VECTOR quaternion from the IMU into a rotation matrix, and applying that to the [y,x,z] coordinates given by YoloSpatialDetectionNetwork. This transforms from the camera's reference frame to the a frame where either x or y is North, and z is up and down relative to gravity. Using the pyquaternion library, something like:

camera_xyz=np.array([y,x,z]) #perhaps the IMU sensor is on sideways, or I have a minus sign wrong somewhere?
quat=Quaternion(rotation_vector_from_imu)
rot=quat.rotation_matrix
map_xyz=rot@camera_xyz