Electromagnetic Inverse Scattering from a Single Transmitter
CVPR 2026 Highlight

TL;DR:

We propose a fully end-to-end, data-driven framework for Electromagnetic Inverse Scattering Problem. It uses data priors to overcome the information scarcity inherent in sparse transmitter setups. Our method thereby achieves robust performance even with a single transmitter and enables real-time inference with over $70,000\times$ speed up.

Electromagnetic Inverse Scattering Problems (EISP) seek to reconstruct relative permittivity from scattered fields and are fundamental to applications like medical imaging. This inverse process is inherently ill-posed and highly nonlinear, making it particularly challenging, especially under sparse transmitter setups, e.g., with only one transmitter. While recent machine learning-based approaches have shown promising results, they often rely on time-consuming, case-specific optimization and perform poorly under sparse transmitter setups. To address these limitations, we revisit EISP from a data-driven perspective. The scarcity of transmitters leads to an insufficient amount of measured data, which fails to capture adequate physical information for stable inversion. Accordingly, we propose a fully end-to-end and data-driven framework that predicts the relative permittivity of scatterers from measured fields, leveraging data distribution priors to compensate for the incomplete information from sparse measurements. This design enables data-driven training and feed-forward prediction of relative permittivity while maintaining strong robustness to transmitter sparsity. Extensive experiments show that our method outperforms state-of-the-art approaches in reconstruction accuracy and robustness. Notably, we demonstrate, for the first time, high-quality reconstruction from a single transmitter. This work advances practical electromagnetic imaging by providing a new, cost-effective paradigm to inverse scattering.

Overview of our method. Our pipeline is built around an MLP that serves as the inverse solver. Given the scattered field measurements $\mathbf{E}^\text{s}$ from all transmitters and receivers, along with a spatial query $\bf{x}$, the MLP directly predicts the relative permittivity $\hat{\boldsymbol{\epsilon}_r}({\bf{x}})$. To enhance spatial expressiveness, we apply positional encoding $\gamma({\bf{x}})$ to the query position. During training, dashed lines indicate the supervision signals applied.

Qualitative comparison under the single-transmitter setting:

Qualitative comparison under the multiple-transmitter setting:

Qualitative comparison under the single-transmitter setting for 3D reconstruction on 3D MNIST dataset:

Qualitative comparison under the single-transmitter setting for 3D reconstruction on 3D ShapeNet dataset: