pith. sign in

arxiv: 2605.29098 · v1 · pith:DC5ZU7VHnew · submitted 2026-05-27 · 💻 cs.CV

Seeing through boxes: Non-Line-of-Sight 3D Reconstruction from Radar Signals

Jiachen Lu , Hailan Shanbhag , Haitham Al Hassanieh This is my paper

Pith reviewed 2026-06-29 12:51 UTC · model grok-4.3

classification 💻 cs.CV
keywords non-line-of-sight reconstructionRF signalsneural fields3D geometryradarsigned distance fieldline-of-sight priorsgeometry reconstruction
0
0 comments X

The pith

GeRaF 2.0 reconstructs both visible and hidden geometry from radar by feeding known outside surfaces as priors into a neural signed-distance field.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows that RF signals can recover accurate 3D surfaces inside boxes or rooms even though they cannot form sharp images on their own. Existing neural methods produce coarse, unstable, and ambiguous results because they ignore how signals travel from the visible region into the hidden one. The authors add the known outside Line-of-Sight geometry directly into the neural-field model so that the optimization respects physical propagation constraints. This single change stabilizes training and produces consistent zero-level sets for both the visible and occluded surfaces. A reader should care because radar can pass through walls where cameras and lidar fail, yet until now the hidden geometry remained too noisy for practical use.

Core claim

The central claim is that a unified LoS-NLoS neural geometry framework called GeRaF 2.0, by integrating visual priors from the outside Line-of-Sight region into the neural field formulation, models RF propagation from the visible area into the enclosed region and thereby achieves stable training together with physically consistent reconstruction of both visible and hidden geometry from radar signals.

What carries the argument

Integration of visual LoS priors into the neural signed-distance-field formulation to guide RF signal propagation from the LoS region into the NLoS region.

If this is right

  • Stable optimization becomes possible for neural RF reconstruction inside enclosed spaces.
  • Both visible and hidden surfaces can be recovered as accurate zero-level sets of the same signed-distance field.
  • Physical consistency of the reconstructed geometry improves because propagation paths are constrained by the known outside surfaces.
  • The method sets a new state-of-the-art on RF-based 3D geometry benchmarks.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same prior-integration idea could be tested on other penetrating-wave modalities such as sonar or through-wall ultrasound.
  • If approximate rather than exact LoS geometry is supplied, the framework might still reduce ambiguity enough for coarse hidden-object detection.
  • Hybrid camera-radar rigs that first map the exterior and then reconstruct the interior become a practical sensing architecture.

Load-bearing premise

The outside Line-of-Sight geometry must be known accurately enough that feeding it as a prior supplies the physical constraints needed to remove surface ambiguity and stabilize optimization.

What would settle it

Running the same neural-field training on identical RF measurements once with the LoS prior and once without it, then measuring whether the version without the prior still produces stable zero-level sets and lower surface error than the version with the prior.

Figures

Figures reproduced from arXiv: 2605.29098 by Hailan Shanbhag, Haitham Al Hassanieh, Jiachen Lu.

Figure 1
Figure 1. Figure 1: This is the first work which uses line-of-sight modalities [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: a) Vision-trained SDF uses negative values (blue) and positive values (red) to separate the inside and outside. b) RF￾trained ULoS SDF models the scene as a series of nested closed and compact sets. Areas with a strong radio frequency interaction (e.g., carton, metal) (blue) are assigned negative values, weak in￾teractions (e.g., air) regions (red) are assigned positive values. 4. Rendering from Radio Freq… view at source ↗
Figure 4
Figure 4. Figure 4: Overall pipeline of GeRaF 2.0. Top: The vision-pretrained SDF on the outside of the box. Bottom: The training pipeline for RF signals. The pipeline begins with lensless sampling. In the first stage of training, we freeze the Reflectivity Network and use the vision-pretrained SDF to adjust transmittance in the ULoS Lensless Rendering module. In the second stage, we use the vision-pretrained SDF to align the… view at source ↗
Figure 5
Figure 5. Figure 5: Illustration of RSDF alignment. In the LoS region [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Qualitative results between vision-based NeuS [ [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 10
Figure 10. Figure 10: Effect of RSDF alignment during training, from Stage [PITH_FULL_IMAGE:figures/full_fig_p009_10.png] view at source ↗
Figure 8
Figure 8. Figure 8: Ablation study on ULoS Lensless Rendering. We com [PITH_FULL_IMAGE:figures/full_fig_p009_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Ablation study on different values of λRSDF. The first column shows Stage 1 results. The top row visualizes surfaces at the zero-level set of the SDF, while the bottom row shows a cross￾sectional slice of the SDF. Stage 1 Stage 2 iter: 1k Stage 2 iter: 5k Stage 2 iter: 10k 𝑔𝑟 = 0 mm 𝑔𝑟 = 0 mm 𝑔𝑟 = 0 mm 𝑔𝑟 = 0 mm [PITH_FULL_IMAGE:figures/full_fig_p009_9.png] view at source ↗
Figure 11
Figure 11. Figure 11: Novel view synthesis on antenna planes at [PITH_FULL_IMAGE:figures/full_fig_p010_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Ablation study on RSDF alignment. We show both 3D [PITH_FULL_IMAGE:figures/full_fig_p013_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Ablation study on number of scanning planes used in [PITH_FULL_IMAGE:figures/full_fig_p013_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Reconstruction results for non-specular objects. [PITH_FULL_IMAGE:figures/full_fig_p014_14.png] view at source ↗
read the original abstract

Reconstructing object geometry from radio frequency (RF) signals is fundamentally challenging due to the lensless imaging nature of RF sensing, which leads to low spatial resolution and high noise. Unlike light signals, RF signals can penetrate occlusions and thus capture information about hidden scenes. Existing Non-Line-of-Sight (NLoS) 3D neural reconstruction methods can recover coarse surfaces inside enclosed environments but often suffer from unstable optimization, noisy surface geometry, and surface ambiguity, failing to produce accurate zero-level sets from the signed distance field (SDF). These limitations largely stem from neglecting the role of Line-of-Sight (LoS) geometry outside the enclosed region, which provides valuable physical constraints for modeling signal propagation. In this paper, we introduce a Unified LoS and NLoS neural geometry reconstruction framework GeRaF 2.0 that leverages the outside LoS geometry to model and guide RF propagation from the LoS region into the NLoS region. By integrating visual LoS priors into the neural field formulation, GeRaF 2.0 achieves stable training and physically consistent reconstruction of both visible and hidden geometry, setting a new state-of-the-art in RF-based geometry reconstruction.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The paper introduces GeRaF 2.0, a unified LoS and NLoS neural geometry reconstruction framework for recovering 3D object geometry from RF signals. It claims that integrating visual Line-of-Sight (LoS) priors into the neural field formulation models and guides RF propagation, yielding stable training, physically consistent reconstruction of visible and hidden geometry, and new state-of-the-art performance in RF-based 3D reconstruction by resolving surface ambiguity in the signed distance field (SDF).

Significance. If the integration of LoS priors demonstrably supplies the claimed physical constraints and produces accurate SDF zero-level sets, the work could advance RF-based non-line-of-sight imaging by combining modalities to handle occlusions. This would be relevant for applications requiring penetration through barriers, provided the method generalizes beyond idealized conditions.

major comments (2)
  1. [Abstract] Abstract: the central claim that integrating visual LoS priors 'models and guides RF propagation' and yields 'physically consistent reconstruction' is unsupported because the abstract (and by extension the manuscript) supplies no equations, loss formulation, or derivation showing how the prior enters the neural SDF or constrains the zero-level set. This is load-bearing for the claim of resolving surface ambiguity.
  2. [Abstract] Abstract and implied methods: the assumption that outside LoS geometry is accurately known and supplies physical constraints is load-bearing, yet no ablation, sensitivity analysis, or error propagation study addresses realistic measurement error in the visual LoS prior. If even modest geometric error is present, the claimed constraints on RF propagation would be violated, leaving NLoS ambiguity unresolved.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on the abstract and the role of LoS priors. We address each point below. Where the manuscript requires clarification or additional analysis, we will revise accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that integrating visual LoS priors 'models and guides RF propagation' and yields 'physically consistent reconstruction' is unsupported because the abstract (and by extension the manuscript) supplies no equations, loss formulation, or derivation showing how the prior enters the neural SDF or constrains the zero-level set. This is load-bearing for the claim of resolving surface ambiguity.

    Authors: The abstract is intentionally concise and omits equations. Section 3 of the manuscript presents the unified neural SDF formulation, in which the visual LoS geometry is incorporated as a boundary condition that defines the entry points for RF propagation into the NLoS region. This enters the optimization through an additional term in the loss that penalizes SDF values inconsistent with the expected travel times and attenuation derived from the LoS-to-NLoS interface. The zero-level set is thereby constrained to surfaces that satisfy both the RF measurements and the known propagation geometry. We will add a one-sentence pointer to this formulation in the revised abstract. revision: partial

  2. Referee: [Abstract] Abstract and implied methods: the assumption that outside LoS geometry is accurately known and supplies physical constraints is load-bearing, yet no ablation, sensitivity analysis, or error propagation study addresses realistic measurement error in the visual LoS prior. If even modest geometric error is present, the claimed constraints on RF propagation would be violated, leaving NLoS ambiguity unresolved.

    Authors: The current experiments assume noise-free visual LoS geometry obtained from standard depth sensors. No dedicated sensitivity study on LoS measurement error appears in the manuscript. This is a substantive limitation of the presented evaluation. We will add an ablation that injects controlled geometric noise into the LoS prior and reports the resulting degradation in NLoS zero-level set accuracy. revision: yes

Circularity Check

0 steps flagged

No circularity; derivation introduces integration of external priors without self-referential reduction

full rationale

The paper's abstract and description present GeRaF 2.0 as a new framework that integrates known visual LoS geometry as a prior into a neural SDF formulation to guide RF propagation modeling for NLoS reconstruction. No equations, fitted parameters renamed as predictions, self-citations as load-bearing uniqueness theorems, or ansatzes smuggled via prior work are quoted or exhibited that would make the output equivalent to the input by construction. The central step is an empirical integration whose physical consistency is claimed to arise from the added prior, not from tautological redefinition of terms. This is the common case of a self-contained proposal whose independence cannot be reduced without further equations.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Review performed on abstract only; full paper may contain additional parameters or assumptions not visible here. The key domain assumption is that LoS geometry supplies usable physical constraints for NLoS RF propagation.

axioms (1)
  • domain assumption Outside LoS geometry provides valuable physical constraints for modeling RF signal propagation from LoS into NLoS regions.
    Explicitly invoked in the abstract as the reason prior methods fail and the new framework succeeds.

pith-pipeline@v0.9.1-grok · 5750 in / 1352 out tokens · 27433 ms · 2026-06-29T12:51:03.641006+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

59 extracted references · 4 canonical work pages

  1. [1]

    See through walls with wifi! In SIGCOMM, 2013

    Fadel Adib and Dina Katabi. See through walls with wifi! In SIGCOMM, 2013. 1

    2013

  2. [2]

    Multi- sensor depth fusion framework for real-time 3d reconstruc- tion.Ieee Access, 7:136471–136480, 2019

    Muhammad Kashif Ali, Asif Rajput, Muhammad Shahzad, Farhan Khan, Faheem Akhtar, and Anko B ¨orner. Multi- sensor depth fusion framework for real-time 3d reconstruc- tion.Ieee Access, 7:136471–136480, 2019. 3

    2019

  3. [3]

    Fast fourier methods for synthetic aperture radar imaging

    Fredrik Andersson, Randolph Moses, and Frank Natterer. Fast fourier methods for synthetic aperture radar imaging. IEEE Transactions on Aerospace and Electronic Systems,

  4. [4]

    Multi-view 3d surface reconstruction from sar images by inverse rendering, 2025

    Emile Barbier-Renard, Florence Tupin, Nicolas Trouv ´e, and Lo¨ıc Denis. Multi-view 3d surface reconstruction from sar images by inverse rendering, 2025. 2

    2025

  5. [5]

    Radar fields: Frequency-space neural scene represen- tations for fmcw radar

    David Borts, Erich Liang, Tim Broedermann, Andrea Ra- mazzina, Stefanie Walz, Edoardo Palladin, Jipeng Sun, David Brueggemann, Christos Sakaridis, Luc Van Gool, et al. Radar fields: Frequency-space neural scene represen- tations for fmcw radar. InSIGGRAPH, 2024. 1, 2

    2024

  6. [6]

    Nerd: Neural reflectance decomposition from image collections

    Mark Boss, Raphael Braun, Varun Jampani, Jonathan T Bar- ron, Ce Liu, and Hendrik Lensch. Nerd: Neural reflectance decomposition from image collections. InICCV, 2021. 2

    2021

  7. [7]

    Anjun Chen, Xiangyu Wang, Kun Shi, Yuchi Huo, Jiming Chen, and Qi Ye. Towards weather-robust 3d human body reconstruction: Millimeter-wave radar-based dataset, bench- mark, and multi-modal fusion.IEEE Transactions on Cir- cuits and Systems for Video Technology, 2024. 3

    2024

  8. [8]

    Radarsim: Simulating single-chip radar via multimodal neural fields

    Chuhan Chen, Tianshu Huang, Akarsh Prabhakara, Chaithanya Kumar Mummadi, Zhongxiao Cong, Anthony Rowe, Matthew O’Toole, and Deva Ramanan. Radarsim: Simulating single-chip radar via multimodal neural fields. InThirteenth International Conference on 3D Vision. 3

  9. [9]

    Neusg: Neural im- plicit surface reconstruction with 3d gaussian splatting guid- ance.arXiv preprint, 2023

    Hanlin Chen, Chen Li, and Gim Hee Lee. Neusg: Neural im- plicit surface reconstruction with 3d gaussian splatting guid- ance.arXiv preprint, 2023. 2

    2023

  10. [10]

    MMDetection3D: Open- MMLab next-generation platform for general 3D object detection.https://github.com/open- mmlab/ mmdetection3d, 2020

    MMDetection3D Contributors. MMDetection3D: Open- MMLab next-generation platform for general 3D object detection.https://github.com/open- mmlab/ mmdetection3d, 2020. 7

    2020

  11. [11]

    Liquid: A wireless liquid identifier

    Ashutosh Dhekne, Mahanth Gowda, Yixuan Zhao, Haitham Hassanieh, and Romit Roy Choudhury. Liquid: A wireless liquid identifier. InProceedings of the 16th annual interna- tional conference on mobile systems, applications, and ser- vices, pages 442–454, 2018. 2

    2018

  12. [12]

    Non-line-of-sight 3d object reconstruction via mmwave surface normal estimation

    Laura Dodds, Tara Boroushaki, Kaichen Zhou, and Fadel Adib. Non-line-of-sight 3d object reconstruction via mmwave surface normal estimation. InMobiCom, 2025. 2, 5

    2025

  13. [13]

    Toward 3d reconstruc- tion of outdoor scenes using an mmw radar and a monocular vision sensor.Sensors, 15(10):25937–25967, 2015

    Ghina El Natour, Omar Ait-Aider, Raphael Rouveure, Franc ¸ois Berry, and Patrice Faure. Toward 3d reconstruc- tion of outdoor scenes using an mmw radar and a monocular vision sensor.Sensors, 15(10):25937–25967, 2015. 3

    2015

  14. [14]

    Relightable 3d gaussians: Re- alistic point cloud relighting with brdf decomposition and ray tracing

    Jian Gao, Chun Gu, Youtian Lin, Zhihao Li, Hao Zhu, Xun Cao, Li Zhang, and Yao Yao. Relightable 3d gaussians: Re- alistic point cloud relighting with brdf decomposition and ray tracing. InECCV, 2024. 2

    2024

  15. [15]

    Implicit geometric regularization for learning shapes.arXiv preprint arXiv:2002.10099,

    Amos Gropp, Lior Yariv, Niv Haim, Matan Atzmon, and Yaron Lipman. Implicit geometric regularization for learning shapes.arXiv preprint arXiv:2002.10099, 2020. 7

    work page arXiv 2002

  16. [16]

    Irgs: Inter-reflective gaussian splatting with 2d gaus- sian ray tracing

    Chun Gu, Xiaofei Wei, Zixuan Zeng, Yuxuan Yao, and Li Zhang. Irgs: Inter-reflective gaussian splatting with 2d gaus- sian ray tracing. InCVPR, 2025. 2

    2025

  17. [17]

    Through fog high-resolution imag- ing using millimeter wave radar

    Junfeng Guan, Sohrab Madani, Suraj Jog, Saurabh Gupta, and Haitham Hassanieh. Through fog high-resolution imag- ing using millimeter wave radar. InCVPR, 2020. 2

    2020

  18. [18]

    2d gaussian splatting for geometrically ac- curate radiance fields

    Binbin Huang, Zehao Yu, Anpei Chen, Andreas Geiger, and Shenghua Gao. 2d gaussian splatting for geometrically ac- curate radiance fields. InSIGGRAPH, 2024. 2

    2024

  19. [19]

    Dart: Im- plicit doppler tomography for radar novel view synthesis

    Tianshu Huang, John Miller, Akarsh Prabhakara, Tao Jin, Tarana Laroia, Zico Kolter, and Anthony Rowe. Dart: Im- plicit doppler tomography for radar novel view synthesis. In CVPR, 2024. 1, 2

    2024

  20. [20]

    3d object reconstruction with mmwave radars.arXiv preprint arXiv:2504.12348,

    Samah Hussein, Junfeng Guan, Swathi Narashiman, Saurabh Gupta, and Haitham Hassanieh. 3d object reconstruction with mmwave radars.arXiv preprint arXiv:2504.12348,

    work page arXiv

  21. [21]

    Gaussianshader: 3d gaussian splatting with shading functions for reflective surfaces

    Yingwenqi Jiang, Jiadong Tu, Yuan Liu, Xifeng Gao, Xiaox- iao Long, Wenping Wang, and Yuexin Ma. Gaussianshader: 3d gaussian splatting with shading functions for reflective surfaces. InCVPR, 2024. 2

    2024

  22. [22]

    Tensoir: Tensorial inverse rendering

    Haian Jin, Isabella Liu, Peijia Xu, Xiaoshuai Zhang, Song- fang Han, Sai Bi, Xiaowei Zhou, Zexiang Xu, and Hao Su. Tensoir: Tensorial inverse rendering. InCVPR, 2023. 2

    2023

  23. [23]

    3d gaussian splatting for real-time radiance field rendering.ACM Transactions on Graphics (TOG), 2023

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk ¨uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering.ACM Transactions on Graphics (TOG), 2023. 2

    2023

  24. [24]

    Radarsplat: Radar gaussian splatting for high-fidelity data synthesis and 3d reconstruc- tion of autonomous driving scenes

    Pou-Chun Kung, Skanda Harisha, Ram Vasudevan, Aline Eid, and Katherine A Skinner. Radarsplat: Radar gaussian splatting for high-fidelity data synthesis and 3d reconstruc- tion of autonomous driving scenes. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 27596–27606, 2025. 1, 2 11

    2025

  25. [25]

    Enabling visual recognition at radio frequency

    Haowen Lai, Gaoxiang Luo, Yifei Liu, and Mingmin Zhao. Enabling visual recognition at radio frequency. InMobiCom,

  26. [26]

    Gs-ir: 3d gaussian splatting for inverse rendering

    Zhihao Liang, Qi Zhang, Ying Feng, Ying Shan, and Kui Jia. Gs-ir: 3d gaussian splatting for inverse rendering. InCVPR,

  27. [27]

    Rcbevdet: Radar-camera fusion in bird’s eye view for 3d object detection

    Zhiwei Lin, Zhe Liu, Zhongyu Xia, Xinhao Wang, Yong- tao Wang, Shengxiang Qi, Yang Dong, Nan Dong, Le Zhang, and Ce Zhu. Rcbevdet: Radar-camera fusion in bird’s eye view for 3d object detection. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14928–14937, 2024. 3

    2024

  28. [28]

    Wave-based non-line-of-sight imaging using fast fk migra- tion.ACM Transactions on Graphics (ToG), 38(4):1–13,

    David B Lindell, Gordon Wetzstein, and Matthew O’Toole. Wave-based non-line-of-sight imaging using fast fk migra- tion.ACM Transactions on Graphics (ToG), 38(4):1–13,

  29. [29]

    Non-line-of-sight imaging using phasor-field virtual wave optics.Nature, 572(7771): 620–623, 2019

    Xiaochun Liu, Ib ´on Guill ´en, Marco La Manna, Ji Hyun Nam, Syed Azer Reza, Toan Huu Le, Adrian Jarabo, Diego Gutierrez, and Andreas Velten. Non-line-of-sight imaging using phasor-field virtual wave optics.Nature, 572(7771): 620–623, 2019. 2

    2019

  30. [30]

    Nero: Neural geometry and brdf reconstruction of reflective objects from multiview images.ACM Transactions on Graphics (TOG), 2023

    Yuan Liu, Peng Wang, Cheng Lin, Xiaoxiao Long, Jiepeng Wang, Lingjie Liu, Taku Komura, and Wenping Wang. Nero: Neural geometry and brdf reconstruction of reflective objects from multiview images.ACM Transactions on Graphics (TOG), 2023. 2

    2023

  31. [31]

    Geraf: Neural geometry reconstruction from radio frequency signals

    Jiachen Lu, Hailan Shanbhag, and Haitham Hassanieh. Geraf: Neural geometry reconstruction from radio frequency signals. InNeurIPS, 2025. 1, 2, 3, 6, 7, 8, 15

    2025

  32. [32]

    Nerf: Representing scenes as neural radiance fields for view syn- thesis.Communications of the ACM, 2021

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view syn- thesis.Communications of the ACM, 2021. 2, 9

    2021

  33. [33]

    Instant neural graphics primitives with a mul- tiresolution hash encoding.ACM Transactions on Graphics (TOG), 2022

    Thomas M ¨uller, Alex Evans, Christoph Schied, and Alexan- der Keller. Instant neural graphics primitives with a mul- tiresolution hash encoding.ACM Transactions on Graphics (TOG), 2022. 4

    2022

  34. [34]

    Extracting triangular 3d models, materials, and lighting from images

    Jacob Munkberg, Jon Hasselgren, Tianchang Shen, Jun Gao, Wenzheng Chen, Alex Evans, Thomas M¨uller, and Sanja Fi- dler. Extracting triangular 3d models, materials, and lighting from images. InCVPR, 2022. 2

    2022

  35. [35]

    Differentiable volumetric rendering: Learn- ing implicit 3d representations without 3d supervision

    Michael Niemeyer, Lars Mescheder, Michael Oechsle, and Andreas Geiger. Differentiable volumetric rendering: Learn- ing implicit 3d representations without 3d supervision. In CVPR, 2020. 2

    2020

  36. [36]

    Confocal non-line-of-sight imaging based on the light-cone transform.Nature, 555(7696):338–341, 2018

    Matthew O’Toole, David B Lindell, and Gordon Wetzstein. Confocal non-line-of-sight imaging based on the light-cone transform.Nature, 555(7696):338–341, 2018. 2

    2018

  37. [37]

    Principles of modern radar

    Mark A Richards, Jim Scheer, William A Holm, and William L Melvin. Principles of modern radar. 2010. 3, 14

    2010

  38. [38]

    Nerv: Neural reflectance and visibility fields for relighting and view synthesis

    Pratul P Srinivasan, Boyang Deng, Xiuming Zhang, Matthew Tancik, Ben Mildenhall, and Jonathan T Barron. Nerv: Neural reflectance and visibility fields for relighting and view synthesis. InCVPR, 2021. 2

    2021

  39. [39]

    Electromagnetic signal attenuation in con- struction materials

    William C Stone. Electromagnetic signal attenuation in con- struction materials. 1997. 2

    1997

  40. [40]

    3drimr: 3d reconstruction and imaging via mmwave radar based on deep learning

    Yue Sun, Zhuoming Huang, Honggang Zhang, Zhi Cao, and Deqiang Xu. 3drimr: 3d reconstruction and imaging via mmwave radar based on deep learning. InIEEE Interna- tional Performance, Computing, and Communications Con- ference (IPCCC), 2021. 2

    2021

  41. [41]

    3d reconstruction of multiple objects by mmwave radar on uav

    Yue Sun, Zhuoming Huang, Honggang Zhang, and Xiaohui Liang. 3d reconstruction of multiple objects by mmwave radar on uav. In2022 IEEE 19th International Conference on Mobile Ad Hoc and Smart Systems (MASS), 2022

    2022

  42. [42]

    R2p: A deep learning model from mmwave radar to point cloud

    Yue Sun, Honggang Zhang, Zhuoming Huang, and Benyuan Liu. R2p: A deep learning model from mmwave radar to point cloud. InInternational Conference on Artificial Neural Networks, 2022. 2

    2022

  43. [43]

    Spinr: Neural volumetric reconstruction for fmcw radars.arXiv preprint,

    Harshvardhan Takawale and Nirupam Roy. Spinr: Neural volumetric reconstruction for fmcw radars.arXiv preprint,

  44. [44]

    Fast 3d reconstruction of space targets from isar image sequences based on neural network

    Weiyi Tan, Yu Wang, Biao Tian, and Shiyou Xu. Fast 3d reconstruction of space targets from isar image sequences based on neural network. InIEEE 8th International Con- ference on Vision, Image and Signal Processing (ICVISP),

  45. [45]

    Texas instrument awr1843.https://www.ti

    TI Inc. Texas instrument awr1843.https://www.ti. com/product/AWR1843, 2023. 7

    2023

  46. [46]

    and Niantic, Inc

    Toolbox AI, Inc. and Niantic, Inc. Scaniverse: 3d scanner app, 2025. Mobile application for iOS. 7

    2025

  47. [47]

    Recovering three-dimensional shape around a corner using ultrafast time-of-flight imaging.Nature communications, 3 (1):745, 2012

    Andreas Velten, Thomas Willwacher, Otkrist Gupta, Ashok Veeraraghavan, Moungi G Bawendi, and Ramesh Raskar. Recovering three-dimensional shape around a corner using ultrafast time-of-flight imaging.Nature communications, 3 (1):745, 2012. 2

    2012

  48. [48]

    Neus: Learning neural implicit surfaces by volume rendering for multi-view reconstruction

    Peng Wang, Lingjie Liu, Yuan Liu, Christian Theobalt, Taku Komura, and Wenping Wang. Neus: Learning neural implicit surfaces by volume rendering for multi-view reconstruction. InNeurIPS, 2021. 2, 3, 4, 5, 7, 8

    2021

  49. [49]

    Safe for generations to come: Considerations of safety for millimeter waves in wireless communications.IEEE mi- crowave magazine, 2015

    Ting Wu, Theodore S Rappaport, and Christopher M Collins. Safe for generations to come: Considerations of safety for millimeter waves in wireless communications.IEEE mi- crowave magazine, 2015. 1

    2015

  50. [50]

    Mvfusion: Multi-view 3d object detection with semantic-aligned radar and camera fusion.arXiv preprint arXiv:2302.10511, 2023

    Zizhang Wu, Guilian Chen, Yuanzhu Gan, Lei Wang, and Jian Pu. Mvfusion: Multi-view 3d object detection with semantic-aligned radar and camera fusion.arXiv preprint arXiv:2302.10511, 2023. 3

    work page arXiv 2023

  51. [51]

    Zfusion: An effective fuser of camera and 4d radar for 3d object perception in autonomous driving

    Sheng Yang, Tong Zhan, Shichen Qiao, Jicheng Gong, Qing Yang, Jian Wang, and Yanfeng Lu. Zfusion: An effective fuser of camera and 4d radar for 3d object perception in autonomous driving. InProceedings of the Computer Vi- sion and Pattern Recognition Conference, pages 3768–3777,

  52. [52]

    Near-field 2-d sar imaging by millimeter-wave radar for concealed item detec- tion

    Muhammet Emin Yanik and Murat Torlak. Near-field 2-d sar imaging by millimeter-wave radar for concealed item detec- tion. In2019 IEEE radio and Wireless Symposium (RWS),

  53. [53]

    Neilf: Neural incident light field for physically-based ma- terial estimation

    Yao Yao, Jingyang Zhang, Jingbo Liu, Yihang Qu, Tian Fang, David McKinnon, Yanghai Tsin, and Long Quan. Neilf: Neural incident light field for physically-based ma- terial estimation. InECCV, 2022. 2

    2022

  54. [54]

    Reflective gaussian splatting

    Yuxuan Yao, Zixuan Zeng, Chun Gu, Xiatian Zhu, and Li Zhang. Reflective gaussian splatting. InICLR, 2025. 2 12

    2025

  55. [55]

    Multiview neu- ral surface reconstruction by disentangling geometry and ap- pearance

    Lior Yariv, Yoni Kasten, Dror Moran, Meirav Galun, Matan Atzmon, Basri Ronen, and Yaron Lipman. Multiview neu- ral surface reconstruction by disentangling geometry and ap- pearance. InNeurIPS, 2020. 2

    2020

  56. [56]

    V ol- ume rendering of neural implicit surfaces

    Lior Yariv, Jiatao Gu, Yoni Kasten, and Yaron Lipman. V ol- ume rendering of neural implicit surfaces. InNeurIPS, 2021. 2

    2021

  57. [57]

    Sparsefusion3d: Sparse sensor fusion for 3d object detection by radar and camera in environmental per- ception.IEEE Transactions on Intelligent Vehicles, 9(1): 1524–1536, 2023

    Zedong Yu, Weibing Wan, Maiyu Ren, Xiuyuan Zheng, and Zhijun Fang. Sparsefusion3d: Sparse sensor fusion for 3d object detection by radar and camera in environmental per- ception.IEEE Transactions on Intelligent Vehicles, 9(1): 1524–1536, 2023. 3

    2023

  58. [58]

    Bodycompass: Monitoring sleep posture with wireless signals.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 4(2):1–25,

    Shichao Yue, Yuzhe Yang, Hao Wang, Hariharan Rahul, and Dina Katabi. Bodycompass: Monitoring sleep posture with wireless signals.Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 4(2):1–25,

  59. [59]

    Rf4d: Neural radar fields for novel view synthesis in outdoor dynamic scenes.arXiv preprint arXiv:2505.20967, 2025

    Jiarui Zhang, Zhihao Li, Chong Wang, and Bihan Wen. Rf4d: Neural radar fields for novel view synthesis in outdoor dynamic scenes.arXiv preprint arXiv:2505.20967, 2025. 1, 2 A. Additional Experiments A.1. Additional RSDF Alignment Ablation In Fig. 12, we present additional ablation studies on RSDF alignment by comparing SDF slices for different objects. Wi...

    work page arXiv 2025