How to create a configuration for RealSense depth input data?

[cansik]

I have created a simple demo on how to feed realsense depth data into the pose network. The problem is that I do not have a specific configuration for the realsense device and I guess that is the reason why the output currently is useless.

Would be great if you could give me a hint in the right direction, or if it is even worth to look into it. Or do you think I would have to retrain the network on the realsense data itself?

[legan78]

Hi,

2D and 3D estimators were trained with kinect 2 data and axus xtion.
In my experience with intel realsense, the noise is even more problematic
to estimate even 2D poses.
I would recommend to retrain the 3D and possibly the 2D estimator.

Do you have an example of the output of the 2D and 3D pose?