PRBonn / kiss-icp

A LiDAR odometry pipeline that just works
https://www.ipb.uni-bonn.de/wp-content/papercite-data/pdf/vizzo2023ral.pdf
MIT License
1.6k stars 321 forks source link

How to recur the poses of kitti dataset? #333

Closed zcc00210 closed 7 months ago

zcc00210 commented 7 months ago

Hello,

Thank you for sharing your work!

When I tried to recur poses.txt using the kitti 05 dataset, I used the command kiss_icp_pipeline --visualize /home/jy/201/velodyne, and in the /home/jy/201/velodyne directory I only used the .bin format of the kitti 05 dataset. image But the resulting velodyne_poses_kitti.txt is nothing like the poses.txt in the original dataset. Although I use evo_traj kitti results/2024-04-22_11-45-32/velodyne_poses_kitti.txt -p --plot_mod=xy to see that the corresponding trajectories for the poses generated using kiss-icp are correct, why do the 12 numbers in each row not being correct? How can I get the same pose as in the kitti dataset?

Thanks in advance!

nachovizzo commented 7 months ago

First, use the kitti dataloader , second use the kitti dataset. You are using semantic kitti, those poses are from SUMA. If you want to see how the trajectory looks like for kiss icp check the eval directory from this repo