Authors: Jiaxiang Tang, Zhaoxi Chen, Xiaokang Chen, Tengfei Wang, Gang Zeng, Ziwei Liu

Published on: February 07, 2024

Impact Score: 8.3

Arxiv code: Arxiv:2402.05054

Summary

• What is new: Introduction of the Large Multi-View Gaussian Model (LGM) for high-resolution 3D content generation from text or single-view images.
• Why this is important: Existing 3D content creation models are limited by low resolution and intensive computation during training.
• What the research proposes: LGM utilizes multi-view Gaussian features and an asymmetric U-Net backbone for efficient, high-resolution 3D model generation.
• Results: Achieved high fidelity 3D models at increased resolution of 512 within 5 seconds, which is significantly faster than previous methods.

Technical Details

Technological frameworks used: Large Multi-View Gaussian Model (LGM)

Models used: Asymmetric U-Net, Multi-view Diffusion Models

Data used: Text prompts, Single-view images

Potential Impact

3D modeling software developers, video game makers, film industry, virtual reality (VR) content creators

Want to implement this idea in a business?

We have generated a startup concept here: Visionary3D.