HyperDiffusion: Generating Implicit Neural Fields with Weight-Space Diffusion
This addresses the problem of synthesizing complex implicit representations for researchers in 3D/4D generative modeling, offering a novel approach but likely incremental as it adapts diffusion models to a new domain.
The paper tackles the challenge of generative modeling for implicit neural fields, which lack a regular grid structure, by proposing HyperDiffusion, a method that operates directly on MLP weights to synthesize new neural implicit fields, enabling high-fidelity generation of 3D shapes and 4D mesh animations within a unified framework.
Implicit neural fields, typically encoded by a multilayer perceptron (MLP) that maps from coordinates (e.g., xyz) to signals (e.g., signed distances), have shown remarkable promise as a high-fidelity and compact representation. However, the lack of a regular and explicit grid structure also makes it challenging to apply generative modeling directly on implicit neural fields in order to synthesize new data. To this end, we propose HyperDiffusion, a novel approach for unconditional generative modeling of implicit neural fields. HyperDiffusion operates directly on MLP weights and generates new neural implicit fields encoded by synthesized MLP parameters. Specifically, a collection of MLPs is first optimized to faithfully represent individual data samples. Subsequently, a diffusion process is trained in this MLP weight space to model the underlying distribution of neural implicit fields. HyperDiffusion enables diffusion modeling over a implicit, compact, and yet high-fidelity representation of complex signals across 3D shapes and 4D mesh animations within one single unified framework.