3D-LDM: Neural Implicit 3D Shape Generation with Latent Diffusion Models
This addresses the challenge of accurate 3D shape generation for applications in computer graphics and AI, representing an incremental improvement over existing diffusion-based methods.
The paper tackles the problem of generating 3D shapes by proposing a diffusion model for neural implicit representations, enabling diverse and high-quality 3D surface generation with capabilities for image-to-3D and text-to-3D conditioning.
Diffusion models have shown great promise for image generation, beating GANs in terms of generation diversity, with comparable image quality. However, their application to 3D shapes has been limited to point or voxel representations that can in practice not accurately represent a 3D surface. We propose a diffusion model for neural implicit representations of 3D shapes that operates in the latent space of an auto-decoder. This allows us to generate diverse and high quality 3D surfaces. We additionally show that we can condition our model on images or text to enable image-to-3D generation and text-to-3D generation using CLIP embeddings. Furthermore, adding noise to the latent codes of existing shapes allows us to explore shape variations.