Neural Electromagnetic Fields for High-Resolution Material Parameter Reconstruction

TL;DR

NEMF is introduced, a novel framework for dense, non-invasive physical inversion designed to build functional digital twins that moves beyond passive visual replicas, enabling the creation of truly functional and simulatable models of the physical world.

Abstract

Creating functional Digital Twins, simulatable 3D replicas of the real world, is a central challenge in computer vision. Current methods like NeRF produce visually rich but functionally incomplete twins. The key barrier is the lack of underlying material properties (e.g., permittivity, conductivity). Acquiring this information for every point in a scene via non-contact, non-invasive sensing is a primary goal, but it demands solving a notoriously ill-posed physical inversion problem. Standard remote signals, like images and radio frequencies (RF), deeply entangle the unknown geometry, ambient field, and target materials. We introduce NEMF, a novel framework for dense, non-invasive physical inversion designed to build functional digital twins. Our key insight is a systematic disentanglement strategy. NEMF leverages high-fidelity geometry from images as a powerful anchor, which first enables the resolution of the ambient field. By constraining both geometry and field using only non-invasive data, the original ill-posed problem transforms into a well-posed, physics-supervised learning task. This transformation unlocks our core inversion module: a decoder. Guided by ambient RF signals and a differentiable layer incorporating physical reflection models, it learns to explicitly output a continuous, spatially-varying field of the scene's underlying material parameters. We validate our framework on high-fidelity synthetic datasets. Experiments show our non-invasive inversion reconstructs these material maps with high accuracy, and the resulting functional twin enables high-fidelity physical simulation. This advance moves beyond passive visual replicas, enabling the creation of truly functional and simulatable models of the physical world.

Figures (5)

• Figure 1: Overview of NEMF framework showing multi-modal fusion for functional digital twins.
• Figure 2: Motivation. (a) Standard Neural Radiance Fields (NeRF) create photorealistic visual twins, but their parameters are implicit and non-physical. (b) Inverting electromagnetic (EM) parameters from sparse signals alone is a highly ill-posed problem. (c) Our work, NEMF, fuses both modalities (visual geometry + physical signals) to create a functional digital twin with explicit, simulatable material properties.
• Figure 3: Our framework systematically decouples the electromagnetic inverse problem, which is challenging and ill posed, into three distinct stages. (1) Stage 1 reconstructs a geometric scaffold ($G$) with high fidelity. (2) Stage 2 reconstructs the ambient radio field ($f_\theta$). (3) Stage 3 uses the frozen $G$ and $f_\theta$ as priors to perform the final physics supervised inversion for the material properties ($g_\phi$).
• Figure 4: Fresnel Inversion Space. The complex relationship, which is not linear, between reflection and material parameters ($\epsilon_r, \sigma$) highlights the ambiguity of the inverse problem. Leveraging data from multiple frequencies robustly constrains the solution for our $\mathcal{L}_{total}$.
• Figure 5: The three indoor scenes used in our synthetic dataset. We use these geometric models, which possess high fidelity, to generate both the images from multiple views for Stage 1 and the sparse CSI data from ray tracing for Stage 2.