1 paper
Weiyu Li, Xuelin Chen, Jue Wang +1
We target a 3D generative model for general natural scenes that are typically unique and intricate. Lacking the necessary volumes of training data, along with the difficulties of h…