ORB_SLAM2
ORB_SLAM2 copied to clipboard
RGB-D with Azure Kinect
Is there anyone who tries to work with Azure Kinect (New released one)
I'm working with D415 and Azure Kinect
but it is not working well with Azure Kinect
Any help or hint about setting yaml file on Azure Kinect ?
- Both Camera data resized with 640x480
- Azure Kinect using NFOV
- Using RGB-D ORB-SLAM2
- DO I NEED TO CHANGE something with ./example/rgbd_tum ?
- wonder what I need to put it with baseline * fx (can't understand fully about this value )
Happy holidays... Thanks in advance
Have you check this issue by @nonlinear1 on Kinect Azure's github? It seems like for high-res rgbd input, orbslam2 tracking could get lost easily.
Or maybe you could try using Kinect Azure ros driver to publish rgbd images from Kinect Azure node in ROS, and feed it to orbslam2 ROS node.
I am currently trying both approaches, but none seems to work. Maybe we could help each other.
For the camera intrinsic values such as fx, fy, you could try using kinect Azure's k4a_device_get_calibration()
function.
I think you should undistort your picture! I give the issue, and finally, I solve this through undistorting the picture!!! for the baseline*fx ,for baseline any value is ok, but may too large is not ok ,for orbslam2 treats rgbd camera as stereo camera!!!
Dear @nonlinear1, I'm trying to get good results with kinect azure but without success, do you have a sample code that work with kinect Azure or a step by step explanation of the process ? Best regards
Give me your e-mail. I will send you the code for input of kinect for azure!
@nonlinear1 thank you. My email is: [email protected]
Give me your e-mail. I will send you the code for input of kinect for azure!
Sir, can I have that?, please? My email : [email protected]
@nonlinear1 Can you give me the code for the input for azure Kinect? I have run ORB-SLAM2 in offline datasets captured with 720P, but the running speed is too slow. I guess that it may be some errors with the parameter setting.
My email: [email protected] Thank you!
@nonlinear1 [email protected] Could you please send me the code,, thank you very much!