I made these NeRFs (Neural Radiance Fields) using Luma Labs AI. Most of them are images for photogrammetry that I had on the hard drive with the exception of the covered barrels. Where NeRFs excel as scene capture tools is in reflections, as the volume representation changes with your viewpoint through colour and opacity. This means it's possible to capture more in a NeRF than a video or photograph - this makes them great for future desktop reference! You can use a sparse number of images to create a NeRF, and the AI can reconstruct synthetic views of the scene based on the supplied info and smooth out camera paths or even create new ones. I'm excited to see where this can fit in to 3D pipeline!
Drone photos as input, shot on an overcast day so not much reflection visible
The bright areas from the environment are reflected in the glove materials here, about 40 photos used so the smoothness of the orbit is impressive. The floating artifacts seem to be related to depth issues, maybe more photos would fix it
Interesting to see the glove isolated from the environment, and the effect of the view dependency
I can almost feel the roughness of the heavy rust on this one
A tree trunk with splitting bark
Barrels covered with tarps, check out the damp puddle reflections on top
Some algae covered rocks on the beach, almost look like cliffs