introlab / rtabmap

RTAB-Map library and standalone application
https://introlab.github.io/rtabmap
Other
2.8k stars 787 forks source link

use Mynteye camera with Rtabmap #423

Open molinadavid opened 5 years ago

molinadavid commented 5 years ago

Hi Mathieu,

We are trying to use a new camera with Rtab-map Mynteye but are having some performance issues, I will share a database later with the sample data privately later.

If you could give us some pointers it would be very helpful.

Best regards, David

matlabbe commented 5 years ago

Hi David,

Looking at the database, you may set RGBD/NeighborLinkRefining to false as the odometry seems quite good already (you will save computation time). I don't recommend setting Kp/MaxFeatures and Vis/MaxFeatures to 0 (infinity), as a lot of computation time could be used if a lot of features are extracted. RGBD/LoopClosureReextractFeatures should not be used in recent versions if Mem/RawDescriptorsKept is true to avoid extracting again exactly the same features (you will save a lot of computation time). Don't use Rtabmap/TimeThr if you don't know what it does. Prefer optimizing all the steps of rtabmap before using memory management. You can open the database in rtabmap-databaseViewer to see all the statistics from the database (Statistics view). In Timing tab, you can see all processing time used for each step of rtabmap.

I see rtabmap's marker detection is enabled. It requires around 300 ms each frame to compute. For high resolution images, you may use apriltag_ros package directly instead, I think there is a decimation option that help to compute the tag detection faster.

cheers, Mathieu