AI-Powered Real-Time Lighting Control (AI-FOR-LIGHT)
January 1, 2025
·
1 min read
Reframed real-time lighting control as a learning-and-evaluation problem: define objectives (visual plausibility vs. cost), baselines, and measurable criteria. Built a reproducible pipeline connecting model inference to an interactive 3D runtime for controlled experiments and rapid iteration. Ran structured comparisons across settings (scenes/conditions) to understand generalization and failure cases, guiding future research directions.
Authors
Baiyan Li
(he/him)
Computer Science Student
I am broadly interested in generative models for interactive visual media, especially games and animation. I aim to develop and adapt GAN-, VAE-, and diffusion-based architectures for controllable content creation and real-time interaction in complex 3D environments.