Librealsense 3D Object/Face reconstruction

[RMang1]

Hello

I'm quite sure there is no middleware for librealsense, such as the FACE and SCAN 3D middleware that were available for the Realsense SDK v.1 (previous to librealsense).

But I am looking for a developer(s) that might be interested in developing a library for facescan that could replicate the output of the v.1 middleware, but with a better precision.Everybody here knows that the v.1 middleware created models that were very light but also very smoothed, which resulted in a poor definition of the face/body/head/whatever you were scanning with the sr300. And the source code was never released so there is no way of tweaking those smoothing parameters to obtain something better.

My goal now is to find a way to obtain more from the sr300, like in the pictures i am attaching, which are screenshots taken from a proprietary software that supports sr300. The quality is much much higher than the v1 middleware released by Intel back then, and I wish to obtain the same results. If there is no available SDK/middleware (open source or paid) I am also willing to pay for someone who can make this.

Thank you for your understanding and sorry if i posted in the wrong section!

Best

[idata · Accepted solution]

Hi robertomangano,

Thank you for your interest in the Intel RealSense SR300 camera.You may find this guide helpful: https://qiita.com/alfredplpl/items/3ed0d9a40d277a19cf7aYou can also post your question on the Github page since there are many RealSense 3rd party developers.Regards,Alexandra

[RMang1]

Hi.

Thank you for your answer.

Ros is more for face recognition, while not so much for 3d scanning (if I understand correctly). But I will write on GitHub and ask there about someone who can help me.

Thank you so much!!!

[LLiu35]

Hi:

For object reconstruction base on D400 series you could find here, https://www.youtube.com/watch?v=N14Pi6z-MkE LIPScan 3D with RealSense D415 - YouTube

For 3D face recognition you could find here, https://www.youtube.com/watch?v=s3xqUh15MDg LIPSface 3D for Windows - YouTube

Hope useful for you

Cheers,

Sneaky@Luke

[RMang1]

Thank you for your answer.

But this looks like a proprietary software, and not an SDK.

Do you know if these guys plan to release an SDK soon?

[MartyG]

If you would like to do face scanning with the RealSense SDK, your best option at present may be to create a 'point cloud' scan of the face.

Here's an example of face tracking with a point cloud, the OpenCV vision software and a Microsoft Kinect camera.

https://www.youtube.com/watch?v=4fJCAQZpvKk Face Point Cloud Tracking using PCL, OpenCV, and Kinect - YouTube

Point clouds can be saved as a 'ply' format file and then converted into a solid mesh model like the one in your supplied images using 3D modeling software such as Blender or MeshLab. There was a discussion several months ago about doing MeshLab conversion.

[RMang1]

Well the thing is that i don't want to rely on other softwares (meshlab, cloudcompare, etc) like the V1 didn't rely on anything but itself.

I see someone is doing it (i.e. Artec Studio 12 works perfectly like i say, and actually they use realsense v1 to access the streams, not even librealsense)