Research map / 3D vision / Face and shape
Imaging Interiors: An Implicit Solution to Electromagnetic Inverse Scattering Problems
Ziyuan Luo, Boxin Shi, Haoliang Li, Renjie Wan
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| represent relative permittivity as an implicit neural representation optimized in a forward process | EISP solver · Institut Fresnel's database | PSNR: 24.95 -> 27.06 (vs Gs SOM) | Improved |
| use INR forward method instead of strongest baseline on MNIST under 30% noise | EISP solver · MNIST dataset (30% noise) | RRMSE: 0.113 -> 0.044 | Improved |
| use two separate MLPs for permittivity and induced current instead of one (ablation) | Proposed · EISP ablation | RRMSE: 0.053 -> 0.038 | Improved |
| use two MLPs instead of single representation (ablation) | Proposed · EISP ablation | time per iteration: 289 ms -> 117 ms | Improved |
| add random spatial sampling instead of uniform sampling (ablation) | Proposed · flexible resolution | RRMSE/SSIM | Improved |
| add total variation loss on contrast (ablation) | Proposed · EISP supplementary ablation | reconstruction quality | Improved |
| use SIREN backbone instead of basic ReLU MLP (ablation) | Proposed · EISP supplementary ablation | reconstruction quality | No clear effect |
| use INR super-resolution sampling instead of Gs SOM + HAT-L post upscaling | Proposed · flexible resolution | image quality | Improved |
| reduce receivers from 32 to 8 (sparse measurement) | Proposed · sparse measurement | reconstruction quality | No clear effect |
| increase noise from 5% to 30% (MNIST) | Proposed · MNIST dataset | RRMSE: 0.017 -> 0.044 | Hurt |
Closest papers by meaning
- NeISF++: Neural Incident Stokes Field for Polarized Inverse Rendering of Conductors and Dielectrics (CVPR 2025 · similarity 0.83)
- Improving Diffusion-based Inverse Algorithms under Few-Step Constraint via Linear Extrapolation (NeurIPS 2025 · similarity 0.81)
- High-Fidelity Polarimetric Implicit 3D Reconstruction with View-Dependent Physical Representation (AAAI 2025 · similarity 0.81)
- Subsurface Scattering for Gaussian Splatting (NeurIPS 2024 · similarity 0.80)
- Ambient Diffusion Posterior Sampling: Solving Inverse Problems with Diffusion Models Trained on Corrupted Data (ICLR 2025 · similarity 0.80)
- NEAR: Neural Electromagnetic Array Response (ICML 2025 · similarity 0.80)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
