Zihan Zheng

Achieve state-of-the-art dynamic 3D reconstruction from sparse views by intelligently fusing explicit 3D geometry with generative priors, avoiding common pitfalls like structural drift and temporal inconsistency.

Real-time, lightweight image compression is now possible with diffusion models, thanks to a novel architecture that swaps transformers for convolutions and prioritizes compression-focused pre-training.

Ditching the 2D latent grid unlocks 60%+ bitrate reductions in generative video compression by encoding videos into adaptable 1D latent tokens.