In rgbd_sequence, when tpose is not matched, why R, t are set to identity matrix?

[xy-guo]

deeptam/tracking/python/deeptam_tracker/evaluation/rgbd_sequence.py

Lines 197 to 206 in e0b7df5

	if tpose:
	timestamp_pose = [tpose] + self.groundtruth_dict[tpose]
	T = transform44(timestamp_pose)
	T = np.linalg.inv(T) # convert to world to cam
	R = T[:3,:3]
	t = T[:3,3]
	else:
	R = np.eye(3)
	t = np.array([0,0,0],dtype=np.float)

Dear authors,

In rgbd_sequence, when tpose is not matched, why R, t are set to identity matrix?

Thanks

[benjaminum]

This is a workaround to make the class work for the sequences without public ground truth.
For sequences with ground truth tpose should always be matched.

[xy-guo]

For some sequences such as fr1/desk2, there are actually several frames (9 out of 640 for fr1/desk2) not matched with ground-truth poses. This causes large evaluation errors.