search

MichaelGrupp / evo

Python package for the evaluation of odometry and SLAM
https://michaelgrupp.github.io/evo/
GNU General Public License v3.0
3.41k stars 746 forks source link

RPE result has something wrong on TUM dataset #260

Closed shelumielzhang closed 4 years ago

shelumielzhang commented 4 years ago

Description: I just want to evaluate my result on TUM dataset, but the rpe result looks something wrong(red circle) rpe Command:

evo_rpe tum groundtruth.txt results.txt -p -a

Console output:


PE w.r.t. translation part (m)
for delta = 1 (frames) using consecutive pairs
(with SE(3) Umeyama alignment)

       max  0.419719
      mean  0.007207
    median  0.005317
       min  0.000560
      rmse  0.013662
       sse  0.308163
       std  0.011606
MichaelGrupp commented 4 years ago

Please check start and end timestamp of your input data. Are they sorted?

shelumielzhang commented 4 years ago

Please check start and end timestamp of your input data. Are they sorted?

Yes, I have checked the timestamp, and here is my input file results.txt

MichaelGrupp commented 4 years ago

No, you have 2 trajectories in the same file.

Look at this check and the plots

evo_traj tum results.txt --full_check --plot
MichaelGrupp commented 4 years ago

It should be fine if you fix your file.

shelumielzhang commented 4 years ago

No, you have 2 trajectories in the same file.

Look at this check and the plots

evo_traj tum results.txt --full_check --plot

Oh, thank you so much! I just use associate.py to align rgb and depth, but why there are two same sequenses in associated.txt?

MichaelGrupp commented 4 years ago

I don't know and I can't help you with this, you need to debug this yourself.