Rendering, light transport and a few sprinkles of physically based simulation. Senior Research Scientist at NVIDIA.
Views and opinions are my own, not NVIDIA's.
@nothings I'm curious, does GPT-4 do better at reversing words? I found Bard and GPT3.5 will output straight gibberish if you ask it to reverse words it hasn't seen in reverse before (and will then lie about it). Does GPT-4 do better than memorizing?
All scenes on the rendering resources page have been updated to support Mitsuba 3. Big thanks to @seb_spe for doing the work!
Grab the scenes here:
https://t.co/1AcrpnikQN
*Mitsuba 3* is now available! It's a major redesign of the lab's infrastructure for differentiable rendering building on the Dr.Jit just-in-time compiler announced yesterday. Full video link: https://t.co/WnfnRWiICm 1/10
Looking forward to (virtually) presenting our work with @ejdeon about accelerating microfacet multiple scattering at EGSR later today!
https://t.co/JLAsz1577H
@BenKrasnow When processing glass on a mill/lathe, is there concern about glass dust/slurry going into the ways? Do you have tricks to deal with that or is it a non-issue?
@yiningkarlli@tylermorganwall Wow! I knew phone lenses did weird stuff, but that's incredible. I really wonder how they even manufacture lenses like this - is this some kind of injection molded plastic, or is this a wholly different exotic manufacturing process I've never heard of?
@BrianKaris Super interesting! Same fidelity than NeRF, much faster training, but much poorer fidelity per bit (~500MB vs 5MB). Neural networks are not magic, but they are data efficient basis functions in this case (i.e. natural scenes).
It's done! I successfully passed my PhD defense today.
A big thanks to my thesis committee, @wkjarosz, @BoZhu88252235, @jaakkolehtinen, Steve Marschner and Peter Winkler for committing their time, and their kind comments.
Onward to new things!