I am using 3 Intel RealSense D435i Depth sensors all positioned on a vertical tripod like shown in the tutorials with the kinects with a turntable for the person to be on, so we will have the person turning at 100/110 cm from the cameras.
the camera’s are running the latest firmware and they have been calibrated in the same spot via the “RealSense View”. I’ve tried multiple sensor configuration in RecFusion Pro and calibrated the system multiple times but the result does not turn quite well as the tutorials.
How can i run a smooth 3d scan without the camera losing depth and stop reconstructing once the subject turns backwards?
A proper multi-sensor calibration is in place. Check if the point clouds align correctly after calibration. You can share screenshots with us for review.
Sufficient geometric detail is available for accurate tracking. You can adjust the volume resolution, but this parameter can be tricky. If the resolution is too high and your computer isn’t fast enough, tracking may be lost. If the resolution is too low, there may not be enough geometric detail for effective tracking.
Consider adding additional objects to the scene to improve tracking. These can be removed after scanning.
If none of these solutions work, feel free to share some sequences with us, and we will review the data.
