search

raulmur / evaluate_ate_scale

Modified tool of the TUM RGB-D dataset that automatically computes the optimal scale factor that aligns trajectory and groundtruth. Useful to evaluate monocular VO/SLAM.
154 stars 65 forks source link

Evaluating Monocular ORB-SLAM/ORB-SLAM2 on KITTI dataset #1

Open gs14iitbbs opened 6 years ago

gs14iitbbs commented 6 years ago

Hi, I want to evaluate monocular ORB-SLAM or Monocular ORB-SLAM2 on KITTI driving dataset. But the format of KITTI groundtruth is

timestamp a b c d e f g h i j k l

, while monocular ORB generates poses in TUM format

timestamp tx ty tz qx qy qz qw

I found this evaluate_rpe and evaluate_ate, which needs both to be in TUM format.

I tried converting keyframe poses from KITTI to TUM(quaternion) format, but got weird results. Can you please explain how to convert KITTI groundtruth poses to TUM format for such evaluation?

Also, the timestamps generated by monocular ORB-SLAM2 are very different from KITTI groundtruth timestamps. I don't understand why?

zssjh commented 6 years ago

@gs14iitbbs did you solved it?

langelesl commented 5 years ago

@gs14iitbbs @zssjh Same question. Have you found a way to evaluate the KeyFrameTrajectory.txt from Mono-Mode with the ground truth file of KITTI? Thanks.

haliphinx commented 5 years ago

@gs14iitbbs @zssjh @langelesl I coded a brief python code to evaluate the mono ORB SLAM on KITTI based on this evaluate_ate_scale file. You can find my code here ORB_MONO_EVA.

joshfriesen commented 2 years ago

@haliphinx your code works wonderfully, thank you.