Kinfu‘s performance is poor.

[star-ice]

• Before opening a new issue, we wanted to provide you with some useful suggestions (Click "Preview" above for a better view):

  • Consider checking out SDK examples.
  • Have you looked in our documentations?
  • Is you question a frequently asked one?
  • Try searching our GitHub Issues (open and closed) for a similar issue.

• All users are welcomed to report bugs, ask questions, suggest or request enhancements and generally feel free to open new issue, even if they haven't followed any of the suggestions above :)

---

Required Info
Camera Model	{ D400 }
Firmware Version	(Open RealSense Viewer --> Click info)
Operating System & Version	Linux(20.04)
Kernel Version (Linux Only)	(e.g. 4.14.13)
Platform	NIVDIA Jetson Orin NX
SDK Version
Language	C++
Segment	Robot

Issue Description

I used Kinfu for the 3D reconstruction of a small scene.
The camera is D455.
I opened the rs-kinfu program and slowly moved the camera, but the performance was not good enough.
The point cloud was skewed and even though I had moved as slowly as I could, there were still so many redundant points around the object.
The point cloud is as follows:

The scene looks like this:

I wonder whether I used kinfu incorrectly or need to modify the code for better performance.
Thanks for your help.

[MartyG-RealSense]

Hi @star-ice It sounds as though you used rs-kinfu properly. Other RealSense users have experienced slow performance with rs-kinfu, so it is not just you.

I would recommend trying the following changes:

1. Comment out line 204 defining the high_density camera configuration preset.

2. Comment out lines 219 and 257-259 that apply post-processing filters, as these filters are calculated on the computer's CPU and so can add a processing burden to it. If commenting out all the filters results in a noticable reduction in depth image quality then try only commenting out the spatial filter.

[star-ice]

It looks quite normal now. Thanks.
I still have another question. Can I use bag file as the input for offline reconstruction? Is there any suggestion for modifing the code?

[MartyG-RealSense]

Yes, you can use a bag file as the data source for a RealSense program script instead of a live camera. You can do this in C++ by inserting a cfg.enable_device_from_file() instruction immediately before the pipe start line.

rs2::config cfg;
cfg.enable_device_from_file("test.bag");
pipe.start(cfg); // Load from file

[star-ice]

Thank you very much @MartyG-RealSense .