search

lukasvst / dm-vio

Source code for the paper DM-VIO: Delayed Marginalization Visual-Inertial Odometry
GNU General Public License v3.0
1.07k stars 187 forks source link

Configuration for EuRoC dataset #5

Closed hellovuong closed 2 years ago

hellovuong commented 2 years ago

Hello, Thank you for the fantastic works. I am able to obtain the very precise result in TUM-VI dataset. However, I got struggle on reproduced the result on EuRoC, sequence MH_04 for specific, (also in OpenLoris datasets), the scale seems bad. I generated the config and run by the python scripts. Do you have any suggest to run it correctly? Thank you in advance!

lukasvst commented 2 years ago

Thanks for trying out the code! Could you please send me the folder with the results so I can have a look later today (the name should start with dmvioresult)? (The folder is generated by the python tools and also contains a file setup/setup.yaml which I can use to see which commands were used to run it.)

hellovuong commented 2 years ago

Thanks for the quick reply. Here is the result folder generated by the scripts

lukasvst commented 2 years ago

Thanks for sending the results! The configuration looks correct but I'll be only able to evaluate the results in a couple of hours as I'm on my phone right now.

Could you do a full run (with 10 iterations) to see if the difference is up to randomness? As the method is non-deterministic, individual runs can differ a bit, but in my experience the results on EuRoC are quite stable overall.

hellovuong commented 2 years ago

Sure, I will run 10 iterations now. But I also think It will not make much difference too :))

hellovuong commented 2 years ago

Close for inactivity

hellovuong commented 2 years ago

Close for inactivity