MichaelGrupp / evo

Python package for the evaluation of odometry and SLAM
https://michaelgrupp.github.io/evo/
GNU General Public License v3.0
3.46k stars 753 forks source link

How to get different sequences of KITTI datasets(KITTI01, .. , 10) #1

Closed haidela closed 6 years ago

haidela commented 6 years ago

Hello, Based on this explanation https://github.com/raulmur/ORB_SLAM2/issues/454 I am trying to do sequences KITTI01,02,.....,10 for S-PTAM, ORB_SLAM and LSD-SLAM 2 days ago unsuccessfully @MichaelGrupp Can you help me to explain the procedure? Thank you in advance for your answer.

MichaelGrupp commented 6 years ago

Please ask more specific what doesn't work for you, what you tried already etc.

haidela commented 6 years ago

@MichaelGrupp I am trying to do KITTI_01_SPTAM.txt, KITTI_02_ORB.txt, KITTI_01_LSD_SLAM.txt(other different sequences (01, 02, ..., 10) of the sequence 00 of KITTI dataset), I don't understand your explanation:Since you want to use TUM RGB-D, KITTI and EuRoC datasets keep in mind that the "default" ground truth trajectory file formats of these datasets are different: TUM: timestamp + xyz + quaternion per row KITTI: flattened SE(3) pose matrix per row (without bottom row 0 0 0 1) EuRoC: csv with state information in IMU/body frame Please, can you explain a little the procedure ?

MichaelGrupp commented 6 years ago

KITTI_01_SPTAM.txt, KITTI_02_ORB.txt, KITTI_01_LSD_SLAM.txt

What are those files? Which format? And what exactly do you want to achieve?

My explanation was just a reminder that there are different trajectory file formats for the ground truth / reference trajectory depending on the dataset.

haidela commented 6 years ago

@MichaelGrupp Hello, I want to make trajectories( in text file format like yours KITTI_00_Orb.text, KITTI_00_SPTAM) performed by S-PTAM and Orb_Slam on KITTI diferent sequences of datasets with my camera parameters. My test steps for ORB_SLAM(based on this issue https://github.com/raulmur/ORB_SLAM/issues/28:

  1. roscore
  2. roslaunch ExampleGroovyOrNewer.launch
  3. rosbag play --pause data/sequence01.bag
  4. python lsd_to_readable.py time.txt calib.txt (time.txt and calib.txt performed by the sequence of KITTI dataset)

SUMMARY

PARAMETERS

$ rostopic list /ORB_SLAM/Frame /ORB_SLAM/Map /ORB_SLAM/Map_array /camera/image_raw /clicked_point /clock /image_view/output /image_view/parameter_descriptions /image_view/parameter_updates /initialpose /move_base_simple/goal /rosout /rosout_agg /tf /tf_static

I didn't successfully tested. Will you explain your procedure to me?

haidela commented 6 years ago

To get the trajectories in text file format

MichaelGrupp commented 6 years ago
  1. Your question is about using the implementations of SLAM algorithms that are not maintained by me and aren't part of this repository. If you need help with running one of those, please open an issue in the repository of the SLAM algorithm you want to use. This will also increase the chances that other users see it who might be able to help you.

Please understand that I only answer specific issues with my software here.

In other words: after you generated poses/trajectories in KITTI, TUM or ROS bag format and you want to use them with this package.

  1. One tip: looking at your infos, it looks like you're using the "old" monocular-only implementation of ORB-SLAM (this one). ORB-SLAM2 is the updated version for monocular, RGB-D and stereo cameras - see here. The README also explains how to use the KITTI dataset. The pose file you will get from this example (will be saved as FrameTrajectory_KITTI_Format.txt) can be compared to the ground truth (download here) using my evo package as described in my README here.
haidela commented 6 years ago

Thank you very much for your valued assistance and your quick reply, which helped me to understand many things

MichaelGrupp commented 6 years ago

You're welcome.