Lightning Talk: Beyond Polygons: The Future of 3D Graphics with Implicit Models
summary
Explains how photogrammetry rigs with about 70–75 cameras feed differentiable inverse rendering that trains volumetric models via gradient descent, and contrasts NeRF (continuous density + view‑dependent color that already works for short dynamic sequences) with Gaussian splatting (discrete “egg” primitives that reconstruct scenes but fail when new geometry appears over time, causing flicker). Claims real‑time production pipelines can move from millions of triangles (Nanite) to implicit fields, but 4D animation remains an engineering bottleneck.
AI-written summary · about
chapters (8)
- 0:003D graphics and implicit models
- 1:06History of 3D asset creation
- 3:59Ray tracing in modern graphics
- 5:483D scanning in film
- 8:15Neural Radiance Fields
- 11:12Gaussian splatting applications
- 13:40Ethics in AI 3D modeling
- 15:05Future 3D content demand
related talks
18mLightning Talk: Turning the Internet Into 3D: 3D Content with NeRF and Gaussian Splat...
37mMake it Blend: Bringing VFX Techniques into the Browser
31mWelcome to 3D
50mHow Neural Networks Helped Graphics
32mMaking pretty things with code: Getting started as Technical Artist
21mPhotorealistic Real Time Computer Graphics with Javascript
42mDeep Learning approaches meet 3D data
01h 03mIntro to 3D Graphics