I have tried using Zed2 sensor to capture images and point cloud depth data. But even after getting flat surface in the view. there is presence of lots of noise in the Point cloud data. This happened in other captures too. Do we have any other way to capture or we need to calibrate the sensor before taking captures? As of now , we have mounted the sensor on a fixed location and angle to start taking captures. Suggestions are welcome.
Welcome to the Stereolabs community
The picture that you attached shows a challenging condition for passive stereo camera processing, with homogeneously colored surfaces and a reflective floor, but our devices can provide good results even in such conditions.
What configuration are you using for the depth estimation? Have you tried the NEURAL depth mode?
We have just tried the default depth mode till now, which seems to be ultra mode. We have to try neural mode as well as other modes too.