You are currently viewing Scene Dreamer

Scene Dreamer

What do you think about this tool?

SceneDreamer is a generative model for unbounded 3D sceneries that can generate large-scale 3D landscapes from random noises. The framework is trained using real-world 2D image collections with no 3D annotations. A principled learning paradigm combining an efficient and expressive 3D scene representation, generative scene parameterization, and an effective renderer that utilizes information from 2D photos is at the heart of the tool. SceneDreamer employs a simplex noise-generated efficient bird’s-eye-view (BEV) representation comprised of a height field and a semantic field.


Visit website

Spread the love

Leave a Reply