Neural Graphics for Improving RealityCapture?

Development Asset Creation
,
1

Hi, I just wanted to know if the latest improvements in neural graphics where the results seem to be available in seconds can somehow allow this to be integrated with RealityCapture to improve results, especially for capturing correctly reflective and transparent surfaces where currently photogrammetry is lacking. This seems to be a fast moving area of research so hoping for some eventual breakthrough :slight_smile: Any thoughts on this? Thanks.

You can see the latest paper and video here:

https://nvlabs.github.io/instant-ngp/

https://nvlabs.github.io/instant-ngp/assets/mueller2022instant.mp4

As well as a simple to understand overall explainer video here:

2

Hi Antony, I added your suggestion also to our feature request database. I deleted the previous post.

3

Hi Ondrej,

Great appreciated, hopefully there is some value in the above neural graphics processing method to improve end results with RealityCapture, I would be interested to hear if that is the case especially given the latest improvements in speed for obtaining such results, thanks. 

4

You’re welcome.