search

ethz-asl / segmap

A map representation based on 3D segments
BSD 3-Clause "New" or "Revised" License
1.06k stars 394 forks source link

Is segmatch deterministic? #55

Closed Tacoma closed 6 years ago

Tacoma commented 6 years ago

Hi all, I have tried the loop closure demonstration and also some own datasets and noticed that different runs with exactly same parameters return different outcomes. I have made small changes to the code so I cannot determine if I introduced the non-deterministic behaviour. Did you also notice non-deterministic behaviour in your experiments and have an idea where it comes from?

vkee commented 6 years ago

Yes, I have gotten that before. I think it is because SegMatch is multi-threaded.

Tacoma commented 6 years ago

Thanks @vkee, that was also my guess besides the random forest and the ROS message system. Good to know that I didn't break it :)

rdube commented 6 years ago

Hi @Tacoma The segmatch code can run single-threaded however the full SLAM integration is multi-threaded (eg. for generating the publishing the local map). It can also be caused by an issue that we experienced with the geometric verification pipeline of the PCL library which we commented on in our IROS17 workshop report. I can send you a link to this once it is online. I hope this helps and happy to hear that the code did not break with your modifications! :)