LGM 3D

What is LGM 3D?

LGM 3D, also known as Large Multi-View Gaussian Model, is a cutting-edge framework designed for high-resolution 3D content creation. It focuses on generating detailed 3D models efficiently from text prompts or single-view images. The innovation lies in its use of multi-view Gaussian features for effective representation and an asymmetric U-Net backbone for high-throughput processing of multi-view images. This approach allows for the generation of high-fidelity 3D objects within a remarkable timeframe of under 5 seconds, while significantly enhancing the training resolution to 512. LGM 3D showcases significant advancements in the field of 3D content creation, balancing both speed and quality effectively.

Who created LGM 3D?

LGM 3D was created by a team of researchers including Jiaxiang Tang, Zhaoxi Chen, Xiaokang Chen, Tengfei Wang, Gang Zeng, and Ziwei Liu. Jiaxiang Tang and Xiaokang Chen are affiliated with Peking University, while Zhaoxi Chen and Ziwei Liu are associated with S-Lab, Nanyang Technological University, and Tengfei Wang is from Shanghai AI Lab . The LGM framework focuses on generating high-resolution 3D models efficiently from text prompts or single-view images, showcasing significant advancements in 3D content creation .

How to use LGM 3D?

To use LGM 3D, follow these steps:

1. Input: Provide text prompts or single-view images to generate 3D models using the LGM framework.

2. Representation: LGM utilizes multi-view Gaussian features for efficient and powerful 3D representation. These features are fused for differentiable rendering.

3. Backbone: The framework employs an asymmetric U-Net as a high-throughput backbone for processing multi-view images. This model can generate multi-view images from text or single-view image inputs via multi-view diffusion models.

4. Speed and Resolution: LGM maintains a high speed, generating 3D objects within 5 seconds, while achieving high-resolution outputs at 512 training resolution.

By following these steps, users can effectively utilize the LGM 3D tool to create high-resolution 3D models from text prompts or single-view images efficiently.