How fast are state-of-the-art computer vision models? Training time might be long, but I wonder if they are fast enough to keep up with high quality real time renderings.
micahtyong
I'm wondering this as well. I think ML-based estimation frameworks like NeRF would do a good job approximating 3D scenes for VR (and other applications) in "real time". NVIDIA worked on a project that could achieve accurate, and essentially real-time performance (e.g., less than a second to render new frames). I found the analysis in this Two Minute Papers video quite relevant to your question!
How fast are state-of-the-art computer vision models? Training time might be long, but I wonder if they are fast enough to keep up with high quality real time renderings.
I'm wondering this as well. I think ML-based estimation frameworks like NeRF would do a good job approximating 3D scenes for VR (and other applications) in "real time". NVIDIA worked on a project that could achieve accurate, and essentially real-time performance (e.g., less than a second to render new frames). I found the analysis in this Two Minute Papers video quite relevant to your question!