LGM 3D

verifiedFreekiui.moe

Open ECCV 2024 research model that generates high-resolution 3D objects from text or a single image in seconds.

What it does

LGM (Large Multi-View Gaussian Model) is an academic research project from Peking University and NTU, published at ECCV 2024. It generates high-resolution 3D content from text prompts or a single image using multi-view Gaussian features and an asymmetric U-Net backbone, producing 3D objects in about five seconds. Code and a demo are publicly available.

Core features

Text-to-3D generation

Single-image-to-3D generation

Multi-view Gaussian representation

Fast (~5s) high-resolution output

Exportable meshes

Open-source code and demo

Best for

→Rapidly prototyping 3D assets

→Turning a single photo into a 3D model

→Research on 3D generative models