DSP-SLAM icon indicating copy to clipboard operation
DSP-SLAM copied to clipboard

Published 20 hours ago • verified publisher icon JingwenWang95

Stereo-only input support?

Open qinyq opened this issue 2 years ago • 3 comments

Hi, thanks for your nice work and sharing the code! I was quite happy when I see the "stereo streams" are supported in the paper. I could successfully reproduce the result (qualitatively).

However, it seems that the released code does not support stereo-only input, right? The kitti dataset uses stereo and lidar data at the same time. Do I miss anything or if there do have support? Thank you!

qinyq avatar Apr 07 '22 10:04 qinyq

Hi, thanks for your interest! Unfortunately we didn't release the stereo-only code because the result was not so good. You can also refer to #11

JingwenWang95 avatar Apr 07 '22 16:04 JingwenWang95

Hi, thanks for your interest! Unfortunately we didn't release the stereo-only code because the result was not so good. You can also refer to #11

Thanks! It's a pity to hear that the stereo-only not working well. Have you got any idea of why? Is it because finding corresponding points? or triangulation?

qinyq avatar Apr 08 '22 02:04 qinyq

Hi, thanks for your interest! Unfortunately we didn't release the stereo-only code because the result was not so good. You can also refer to #11

Thanks! It's a pity to hear that the stereo-only not working well. Have you got any idea of why? Is it because finding corresponding points? or triangulation?

Actually there are several reasons, but we didn't have enough time to conduct more thourough analysis to try to solve this. But the two major issues I observed were: 1. the initial pose estimate is way worse. 2. the SLAM points are way noisier. I'm happy to have further discussions and you can reach out to me via email.

JingwenWang95 avatar Apr 10 '22 17:04 JingwenWang95