Enhancement of the precision of map-db

[sabosabo77]

What I Want to Achieve

I want to enhance the precision of the map database used in Localization mode.

Input:
• Camera trajectory (high-accuracy self-positioning results, such as those from LIDAR-SLAM)
• Monocular RGB images at each timestamp

Output:
• Map database

Question

Even with offline processing, I believe this can be achieved by calling some of the functions already implemented in Stella-VSLAM from my custom program. However, the code is complex, and I’m struggling to understand the behavior of each function.

Could you provide general guidance on which functions to call and in what order?
Additionally, if there are any documents that can help me understand Stella-VSLAM’s algorithm, I would appreciate it if you could share them.

[Skwangles]

Hi,

Can you please expand on what you mean by 'enhance precision'?
Do you mean getting better localisation accuracy, or something - I'm not the best in that area, but I can help with codebase questions.

I'm not too sure what you mean about improving the 'map database' precision, the only thing I can think of is the decimal precision - you can get the full precision keyframe positions by using the Publisher interface the examples use. try seeing Iridescence viewer.

Otherwise, I'm not too sure how to help

Here's an answer to your 'Additionally' request:
OpenVSLAM (Stella VSLAM's original project) t is a reimplementation of the OrbSLAM2 paper.
Here is the actual OpenVSLAM paper that Stella VSLAM has been forked from.