pith. sign in

arxiv: 2604.22827 · v1 · submitted 2026-04-19 · 💻 cs.CV · cs.LG

DGHMesh: A Large-scale Dual-radar mmWave Dataset and Generalization-focused Benchmark for Human Mesh Reconstruction

Rongxiao Guo , Qingchao Chen This is my paper

Pith reviewed 2026-05-10 05:56 UTC · model grok-4.3

classification 💻 cs.CV cs.LG
keywords mmWave radarhuman mesh reconstructiondual-radar datasetgeneralization benchmarkmulti-radar fusionpoint cloudsimaging tubesFMCW radar
0
0 comments X

The pith

Fusing point clouds and imaging tubes from dual mmWave radars yields accurate human mesh reconstruction that generalizes across shifts in position, orientation, and subject.

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

This paper introduces DGHMesh, a dataset of 360,000 synchronized frames from 15 subjects performing 8 actions, captured with both FMCW and SFCW mmWave radars along with RGB images and precise 3D mesh ground truth. It defines a benchmark that measures reconstruction performance when subjects change location, face direction, radar subarray sizes, or when the test person is unseen during training. The authors also propose mmPTM, a query-based model that fuses point clouds and imaging tubes from the two radars. Experiments across the benchmark settings show that this fusion produces higher accuracy than baselines while preserving competitive performance under the tested changes, which supports development of more reliable contactless sensing systems.

Core claim

DGHMesh supplies synchronized raw I/Q data from FMCW and SFCW radars, calibrated spatial positions, and 3D human mesh annotations across 360,000 frames from 15 subjects and 8 actions. The generalization benchmark consists of sub-tasks for human position shifts, orientation shifts, subarray size variations, and cross-subject evaluation. The mmPTM framework, which uses queries to jointly process multi-radar point clouds and imaging tubes, achieves outstanding accuracy and competitive generalization across these sub-benchmarks.

What carries the argument

mmPTM, a query-based multi-radar fusion framework that jointly exploits point clouds and imaging tubes from dual mmWave radars to reconstruct human meshes.

If this is right

  • Human mesh reconstruction algorithms can now be compared fairly on their ability to handle real deployment variations such as moving people or altered radar parameters.
  • Combining point cloud and imaging representations from two radar types produces more stable results than single-radar methods when conditions change.
  • Public release of raw I/Q signals and accurate calibrations enables end-to-end learning pipelines for mmWave-based sensing.
  • Generalization-focused evaluation helps identify approaches suitable for practical privacy-preserving monitoring without per-deployment retraining.

Where Pith is reading between the lines

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

  • If the observed generalization extends further, the method could support mmWave sensing in varied indoor spaces without frequent model updates.
  • The dual-radar fusion pattern may transfer to other wave-based sensing tasks that combine sparse and dense representations.
  • Cross-subject results hint that modest subject diversity in training can yield models adaptable to new users with limited extra data.

Load-bearing premise

The variability captured in data from 15 subjects and 8 actions under the tested configuration shifts is representative enough that reported generalization performance will transfer to unseen subjects, environments, or radar hardware.

What would settle it

A large accuracy drop for mmPTM on new recordings from additional subjects in unfamiliar environments or with different untested radar hardware would show that the generalization results do not extend beyond the benchmark conditions.

Figures

Figures reproduced from arXiv: 2604.22827 by Qingchao Chen, Rongxiao Guo.

Figure 1
Figure 1. Figure 1: System overview. (a) Data acquisition hardware and experimental site; (b) The rear side of the acquisition fixture with reflective markers attached; [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the proposed mmPTM. The FMCW radar point cloud is encoded by the point cloud encoder into point tokens [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Visualizations of synchronized multimodal data in DGHMesh. For each action, the RGB image, 3D mesh annotation, FMCW point cloud, and SFCW [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Action Sequences. The gray grids denote the activity area of the [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Subarray Configuration. Red circles denote the selected Tx, while light [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Visualization of cross-subject evaluation results. The first column shows the ground-truth meshes, while the remaining columns present the meshes [PITH_FULL_IMAGE:figures/full_fig_p012_6.png] view at source ↗
read the original abstract

Millimeter-wave (mmWave) radar has shown great potential for contactless, privacy-preserving, and robust human sensing, yet existing mmWave-based human mesh reconstruction (HMR) studies are still limited by the lack of benchmarks for generalization analysis under configuration shifts and fair comparison of different algorithms. To address the limitation, we present DGHMesh, a large-scale dual-radar mmWave dataset and generalization-focused benchmark for HMR. It contains data from 15 subjects performing 8 actions, with 360,000 synchronized frames collected from FMCW radar, SFCW radar, RGB images, and high-precision 3D HMR annotations. In addition, the dataset provides synchronized raw I/Q data from both radar modalities and accurately calibrated radar spatial positions. The benchmark is designed to evaluate HMR methods under diverse measurement configurations, including human position shifts, human orientation shifts, subarray size variations, and cross-subject settings. Based on DGHMesh, we also propose mmPTM, a query-based multi-radar fusion framework that jointly exploits point clouds and imaging tubes for HMR. Extensive experiments are conducted against representative baselines under different settings. The results demonstrate that mmPTM consistently achieves outstanding accuracy and competitive generalization capability across multiple sub-benchmarks, validating the effectiveness of multi-radar fusion and the practical value of the proposed dataset and benchmark for mmWave-based HMR research. DGHMesh and mmPTM are publicly available at https://github.com/SPIresearch/DGHMesh.(The complete benchmark and code will be released after paper publication)

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

1 major / 2 minor

Summary. The manuscript presents DGHMesh, a dual-radar mmWave dataset with 360,000 synchronized frames collected from 15 subjects performing 8 actions, including raw I/Q data from FMCW and SFCW radars, RGB images, and high-precision 3D HMR annotations. It introduces mmPTM, a query-based multi-radar fusion framework that exploits point clouds and imaging tubes, and establishes a benchmark evaluating HMR methods under configuration shifts (position, orientation, subarray size, cross-subject). The central claim is that mmPTM achieves outstanding accuracy and competitive generalization across sub-benchmarks, demonstrating the value of multi-radar fusion and the proposed dataset.

Significance. If the generalization results hold under scrutiny, the work would make a useful contribution by releasing a public dual-radar mmWave dataset and benchmark focused on configuration shifts, filling a gap in mmWave HMR research where existing studies lack standardized generalization tests. The synchronized raw data and calibrated positions could support further multi-modal work. The empirical nature of the contribution means its impact depends on the robustness of the cross-subject and cross-configuration results.

major comments (1)
  1. The claim of 'competitive generalization capability' (Abstract) across sub-benchmarks, including cross-subject settings, is based on data from only 15 subjects and 8 actions. Human mesh reconstruction is sensitive to inter-subject differences in body shape, proportions, gait style, and clothing; a cohort of this size provides limited coverage of population variability. The tested configuration shifts (position, orientation, subarray size) are intra-subject or hardware-internal and do not substitute for subject diversity, so the reported metrics may not transfer to unseen subjects or real deployments.
minor comments (2)
  1. Abstract: the statement that 'extensive experiments are conducted against representative baselines' provides no quantitative metrics, baseline details, error bars, or data-split descriptions; adding key numbers would make the summary more informative.
  2. Availability statement: the claim that 'DGHMesh and mmPTM are publicly available' is immediately qualified by 'The complete benchmark and code will be released after paper publication,' creating ambiguity about current access.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We have carefully considered the major comment and outline our response and planned revisions below.

read point-by-point responses

  1. Referee: The claim of 'competitive generalization capability' (Abstract) across sub-benchmarks, including cross-subject settings, is based on data from only 15 subjects and 8 actions. Human mesh reconstruction is sensitive to inter-subject differences in body shape, proportions, gait style, and clothing; a cohort of this size provides limited coverage of population variability. The tested configuration shifts (position, orientation, subarray size) are intra-subject or hardware-internal and do not substitute for subject diversity, so the reported metrics may not transfer to unseen subjects or real deployments.

    Authors: We agree that the subject cohort size of 15 and the set of 8 actions represent a genuine limitation for claims of generalization, as human mesh reconstruction performance can vary substantially with inter-subject differences in body shape, proportions, gait, and clothing. Our benchmark does evaluate cross-subject settings independently from the configuration shifts (position, orientation, and subarray size), but we acknowledge that these evaluations are confined to the available cohort and do not fully substitute for broader population diversity or real-world deployment conditions. In the revised manuscript, we will moderate the abstract language from 'competitive generalization capability' to 'promising generalization performance within the evaluated settings' and add an explicit limitations paragraph in the discussion section that addresses subject diversity, its implications for transferability, and directions for future larger-scale data collection. These changes will provide a more accurate and balanced presentation of the results. revision: partial

Circularity Check

0 steps flagged

No significant circularity; empirical dataset and benchmark evaluation

full rationale

The paper introduces a new dual-radar dataset (DGHMesh) collected from 15 subjects and proposes the mmPTM fusion framework, then reports experimental results on accuracy and generalization across sub-benchmarks including cross-subject splits. No mathematical derivation chain, equations, or 'predictions' exist that reduce by construction to fitted parameters, self-defined quantities, or prior self-citations. All performance claims are direct empirical measurements on held-out portions of the collected data; the limited subject count affects external validity but does not create circularity within the reported results.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claims rest on standard domain assumptions in radar signal processing and deep learning for 3D reconstruction; no free parameters, new axioms, or invented entities are introduced beyond those already established in the field.

axioms (1)
  • domain assumption mmWave radar returns can be transformed into point clouds and imaging tubes that contain sufficient geometric information for 3D human mesh reconstruction
    Invoked implicitly as the foundation for both the dataset utility and the mmPTM architecture.

pith-pipeline@v0.9.0 · 5588 in / 1347 out tokens · 44084 ms · 2026-05-10T05:56:46.744776+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

55 extracted references · 55 canonical work pages

  1. [1]

    Millimeter-wave radar for intelligent sensing: A comprehensive review of techniques, applica- tions, and challenges,

    Y . Soni, M. Goswami, N. P. Shetty,et al., “Millimeter-wave radar for intelligent sensing: A comprehensive review of techniques, applica- tions, and challenges,”Computers and Electrical Engineering, vol. 128, p. 110696, 2025

    work page 2025

  2. [2]

    A survey of mmwave radar- based sensing in autonomous vehicles, smart homes and industry,

    H. Kong, C. Huang, J. Yu, and X. Shen, “A survey of mmwave radar- based sensing in autonomous vehicles, smart homes and industry,”IEEE Communications Surveys & Tutorials, vol. 27, no. 1, pp. 463–508, 2024

    work page 2024

  3. [3]

    4d mmwave radar for autonomous driving perception: A comprehensive survey,

    L. Fan, J. Wang, Y . Chang, Y . Li, Y . Wang, and D. Cao, “4d mmwave radar for autonomous driving perception: A comprehensive survey,” IEEE Transactions on Intelligent V ehicles, vol. 9, no. 4, pp. 4606–4620, 2024

    work page 2024

  4. [4]

    Non-contact monitoring of fatigue driving using fmcw millimeter wave radar,

    H. Chen, X. Han, Z. Hao, H. Yan, and J. Yang, “Non-contact monitoring of fatigue driving using fmcw millimeter wave radar,”ACM Transactions on Internet of Things, vol. 5, no. 1, pp. 1–18, 2023

    work page 2023

  5. [5]

    Towards domain-independent and real-time gesture recognition using mmwave signal,

    Y . Li, D. Zhang, J. Chen, J. Wan, D. Zhang, Y . Hu, Q. Sun, and Y . Chen, “Towards domain-independent and real-time gesture recognition using mmwave signal,”IEEE Transactions on Mobile Computing, vol. 22, no. 12, pp. 7355–7369, 2022

    work page 2022

  6. [6]

    Omni- directional human activity recognition with spatial–temporal trajectory- based compensation using monostatic mimo radar,

    C. Ding, J. Weng, H. Zhao, Q. Zhou, H. Hong, and X. Zhu, “Omni- directional human activity recognition with spatial–temporal trajectory- based compensation using monostatic mimo radar,”IEEE Transactions on Microwave Theory and Techniques, 2025

    work page 2025

  7. [7]

    mmmotion: A real- time human motion detection smart home system based on mmwave radar,

    R. Zeng, J. Zhang, Y . Mao, Z. Yang, and L. Shi, “mmmotion: A real- time human motion detection smart home system based on mmwave radar,”IEEE Internet of Things Journal, 2025

    work page 2025

  8. [8]

    Vital signs detection with difference beamforming and orthogonal projection filter based on simo-fmcw radar,

    J. Xiong, H. Hong, L. Xiao, E. Wang, and X. Zhu, “Vital signs detection with difference beamforming and orthogonal projection filter based on simo-fmcw radar,”IEEE Transactions on Microwave Theory and Techniques, vol. 71, no. 1, pp. 83–92, 2022

    work page 2022

  9. [9]

    Non-intrusive human vital sign detection using mmwave sensing technologies: A review,

    Y . Wu, H. Ni, C. Mao, J. Han, and W. Xu, “Non-intrusive human vital sign detection using mmwave sensing technologies: A review,”ACM Transactions on Sensor Networks, vol. 20, no. 1, pp. 1–36, 2023

    work page 2023

  10. [10]

    Radar-based millimeter- wave sensing for accurate 3-d indoor positioning: Potentials and chal- lenges,

    A. Sesyuk, S. Ioannou, and M. Raspopoulos, “Radar-based millimeter- wave sensing for accurate 3-d indoor positioning: Potentials and chal- lenges,”IEEE Journal of Indoor and Seamless Positioning and Naviga- tion, vol. 2, pp. 61–75, 2024

    work page 2024

  11. [11]

    Radar can see and hear as well: A new multimodal benchmark based on radar sensing,

    Y . Xu and Q. Chen, “Radar can see and hear as well: A new multimodal benchmark based on radar sensing,”IEEE Internet of Things Journal, vol. 11, no. 15, pp. 26459–26469, 2024

    work page 2024

  12. [12]

    A comprehensive survey of research trends in mmwave technologies for medical applications,

    X. Zhang, C. Liu, Y . Cheng, Z. Li, C. Xu, C. Huang, Y . Zhan, W. Bo, J. Xia, and W. Xu, “A comprehensive survey of research trends in mmwave technologies for medical applications,”Sensors, vol. 25, no. 12, p. 3706, 2025

    work page 2025

  13. [13]

    mm-fall: Practical and robust fall detection via mmwave signals,

    C. Zhao, Q. Luo, H. Ding, G. Wang, K. Zhao, Z. Wang, W. Xi, and J. Zhao, “mm-fall: Practical and robust fall detection via mmwave signals,”IEEE Transactions on Mobile Computing, 2025

    work page 2025

  14. [14]

    Misleep: Human sleep posture identification from deep learning augmented millimeter-wave wireless systems,

    A. Adhikari and S. Sur, “Misleep: Human sleep posture identification from deep learning augmented millimeter-wave wireless systems,”ACM Transactions on Internet of Things, vol. 5, no. 2, pp. 1–33, 2024

    work page 2024

  15. [15]

    Motion- tolerant measurement of respiration and heartbeat via mmwave radar across diverse healthcare scenarios,

    J. Qin, H. Chen, J. Cheng, X. Liu, X. Zhang, and A. Song, “Motion- tolerant measurement of respiration and heartbeat via mmwave radar across diverse healthcare scenarios,”IEEE Transactions on Instrumen- tation and Measurement, 2026

    work page 2026

  16. [16]

    Non-contact monitoring of human car- diorespiratory activity during sleep using fmcw millimeter wave radar,

    E.-K. Wu, Q.-G. Fan, M.-C. Li, J.-H. Zhang, J. Jia, T. Qiang, C. Wang, X.-F. Gu, and J.-G. Liang, “Non-contact monitoring of human car- diorespiratory activity during sleep using fmcw millimeter wave radar,” Measurement, vol. 242, p. 116144, 2025

    work page 2025

  17. [17]

    A con- current multibeam passive radar for respiratory detection from multiple subjects,

    S. Huang, H. Zhao, H. Hong, F. Guo, M. Peng, and X. Zhu, “A con- current multibeam passive radar for respiratory detection from multiple subjects,”IEEE Transactions on Microwave Theory and Techniques, 2025

    work page 2025

  18. [18]

    A novel miniaturized 15 mmwave antenna sensor for breast tumor detection and 5g communica- tion,

    C. Das, M. Z. Chowdhury, and Y . M. Jang, “A novel miniaturized 15 mmwave antenna sensor for breast tumor detection and 5g communica- tion,”IEEE Access, vol. 10, pp. 114856–114868, 2022

    work page 2022

  19. [19]

    mmarrhythmia: Contactless arrhythmia detection via mmwave sensing,

    L. Zhao, R. Lyu, Q. Lin, A. Zhou, H. Zhang, H. Ma, J. Wang, C. Shao, and Y . Tang, “mmarrhythmia: Contactless arrhythmia detection via mmwave sensing,”Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, vol. 8, no. 1, pp. 1–25, 2024

    work page 2024

  20. [20]

    Bridging data gaps in healthcare: a scoping review of transfer learning in structured data analysis,

    S. Li, X. Li, K. Yu, Q. Wu, D. Miao, M. Zhu, M. Yan, Y . Ke, D. D’Agostino, Y . Ning,et al., “Bridging data gaps in healthcare: a scoping review of transfer learning in structured data analysis,”Health Data Science, vol. 5, p. 0321, 2025

    work page 2025

  21. [21]

    A health monitoring system with posture estimation and heart rate detection based on millimeter-wave radar,

    J. Wu and N. Dahnoun, “A health monitoring system with posture estimation and heart rate detection based on millimeter-wave radar,” Microprocessors and Microsystems, vol. 94, p. 104670, 2022

    work page 2022

  22. [22]

    Toward weather-robust 3d human body reconstruction: Millimeter-wave radar- based dataset, benchmark, and multi-modal fusion,

    A. Chen, X. Wang, K. Shi, Y . Huo, J. Chen, and Q. Ye, “Toward weather-robust 3d human body reconstruction: Millimeter-wave radar- based dataset, benchmark, and multi-modal fusion,”IEEE Transactions on Circuits and Systems for Video Technology, vol. 35, no. 1, pp. 273– 286, 2024

    work page 2024

  23. [23]

    M4human: A large-scale mul- timodal mmwave radar benchmark for human mesh reconstruction.arXiv preprint arXiv:2512.12378, 2025

    J. Fan, Y . Zhou, Y . Yang, X. Cui, J. Zhang, L. Xie, J. Yang, C. X. Lu, and F. Ding, “M4human: A large-scale multimodal mmwave radar bench- mark for human mesh reconstruction,”arXiv preprint arXiv:2512.12378, 2025

    work page arXiv 2025

  24. [24]

    Deep learning-based human pose estimation: A survey,

    C. Zheng, W. Wu, C. Chen, T. Yang, S. Zhu, J. Shen, N. Kehtarnavaz, and M. Shah, “Deep learning-based human pose estimation: A survey,” ACM computing surveys, vol. 56, no. 1, pp. 1–37, 2023

    work page 2023

  25. [25]

    Recovering 3d human mesh from monocular images: A survey,

    Y . Tian, H. Zhang, Y . Liu, and L. Wang, “Recovering 3d human mesh from monocular images: A survey,”IEEE transactions on pattern analysis and machine intelligence, vol. 45, no. 12, pp. 15406–15425, 2023

    work page 2023

  26. [26]

    Human3. 6m: Large scale datasets and predictive methods for 3d human sensing in natural environments,

    C. Ionescu, D. Papava, V . Olaru, and C. Sminchisescu, “Human3. 6m: Large scale datasets and predictive methods for 3d human sensing in natural environments,”IEEE transactions on pattern analysis and machine intelligence, vol. 36, no. 7, pp. 1325–1339, 2013

    work page 2013

  27. [27]

    3d human mesh recovery: Comparative review, models, and prospects,

    W. Kim, “3d human mesh recovery: Comparative review, models, and prospects,”Journal of Visual Communication and Image Representation, vol. 115, p. 104699, 2026

    work page 2026

  28. [28]

    mmmesh: Towards 3d real-time dynamic human mesh construction using millimeter-wave,

    H. Xue, Y . Ju, C. Miao, Y . Wang, S. Wang, A. Zhang, and L. Su, “mmmesh: Towards 3d real-time dynamic human mesh construction using millimeter-wave,” inProceedings of the 19th annual international conference on mobile systems, applications, and services, pp. 269–282, 2021

    work page 2021

  29. [29]

    mmbody benchmark: 3d body reconstruction dataset and analysis for millimeter wave radar,

    A. Chen, X. Wang, S. Zhu, Y . Li, J. Chen, and Q. Ye, “mmbody benchmark: 3d body reconstruction dataset and analysis for millimeter wave radar,” inProceedings of the 30th ACM International Conference on Multimedia, pp. 3501–3510, 2022

    work page 2022

  30. [30]

    Definitions and examples of inverse and ill- posed problems,

    S. I. Kabanikhinet al., “Definitions and examples of inverse and ill- posed problems,”J. Inverse Ill-Posed Probl, vol. 16, no. 4, pp. 317–357, 2008

    work page 2008

  31. [31]

    S. I. Kabanikhin,Inverse and ill-posed problems: theory and applica- tions. de Gruyter, 2011

    work page 2011

  32. [32]

    Rf-based 3d skeletons,

    M. Zhao, Y . Tian, H. Zhao, M. A. Alsheikh, T. Li, R. Hristov, Z. Kabelac, D. Katabi, and A. Torralba, “Rf-based 3d skeletons,” inProceedings of the 2018 conference of the ACM special interest group on data communication, pp. 267–281, 2018

    work page 2018

  33. [33]

    mm-pose: Real-time human skeletal posture estimation using mmwave radars and cnns,

    A. Sengupta, F. Jin, R. Zhang, and S. Cao, “mm-pose: Real-time human skeletal posture estimation using mmwave radars and cnns,”IEEE sensors journal, vol. 20, no. 17, pp. 10032–10044, 2020

    work page 2020

  34. [34]

    Hupr: A benchmark for human pose estimation using millimeter wave radar,

    S.-P. Lee, N. P. Kini, W.-H. Peng, C.-W. Ma, and J.-N. Hwang, “Hupr: A benchmark for human pose estimation using millimeter wave radar,” inProceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pp. 5715–5724, 2023

    work page 2023

  35. [35]

    Mars: mmwave-based assistive rehabilitation system for smart healthcare,

    S. An and U. Y . Ogras, “Mars: mmwave-based assistive rehabilitation system for smart healthcare,”ACM Transactions on Embedded Comput- ing Systems (TECS), vol. 20, no. 5s, pp. 1–22, 2021

    work page 2021

  36. [36]

    mri: Multi-modal 3d human pose estimation dataset using mmwave, rgb-d, and inertial sensors,

    S. An, Y . Li, and U. Ogras, “mri: Multi-modal 3d human pose estimation dataset using mmwave, rgb-d, and inertial sensors,”Advances in neural information processing systems, vol. 35, pp. 27414–27426, 2022

    work page 2022

  37. [37]

    Mm-fi: Multi-modal non-intrusive 4d human dataset for versatile wireless sensing,

    J. Yang, H. Huang, Y . Zhou, X. Chen, Y . Xu, S. Yuan, H. Zou, C. X. Lu, and L. Xie, “Mm-fi: Multi-modal non-intrusive 4d human dataset for versatile wireless sensing,”Advances in Neural Information Processing Systems, vol. 36, pp. 18756–18768, 2023

    work page 2023

  38. [38]

    Rt-pose: A 4d radar tensor-based 3d human pose estimation and localization benchmark,

    Y .-H. Ho, J.-H. Cheng, S. Y . Kuan, Z. Jiang, W. Chai, H.-W. Huang, C.- L. Lin, and J.-N. Hwang, “Rt-pose: A 4d radar tensor-based 3d human pose estimation and localization benchmark,” inEuropean Conference on Computer Vision, pp. 107–125, Springer, 2024

    work page 2024

  39. [39]

    Mm- dcdr: A benchmark of device configuration and data representation for mmwave-based human sensing,

    Y . Gao, R. Geng, D. Zhang, Y . Hu, H. Lin, and Y . Chen, “Mm- dcdr: A benchmark of device configuration and data representation for mmwave-based human sensing,” in2024 16th International Conference on Wireless Communications and Signal Processing (WCSP), pp. 801– 806, IEEE, 2024

    work page 2024

  40. [40]

    Three-dimensional reconstruction from a single rgb image using deep learning: A review,

    M. S. U. Khan, A. Pagani, M. Liwicki, D. Stricker, and M. Z. Afzal, “Three-dimensional reconstruction from a single rgb image using deep learning: A review,”Journal of Imaging, vol. 8, no. 9, p. 225, 2022

    work page 2022

  41. [41]

    Keep it smpl: Automatic estimation of 3d human pose and shape from a single image,

    F. Bogo, A. Kanazawa, C. Lassner, P. Gehler, J. Romero, and M. J. Black, “Keep it smpl: Automatic estimation of 3d human pose and shape from a single image,” inEuropean conference on computer vision, pp. 561–578, Springer, 2016

    work page 2016

  42. [42]

    Convolutional mesh regression for single-image human shape reconstruction,

    N. Kolotouros, G. Pavlakos, and K. Daniilidis, “Convolutional mesh regression for single-image human shape reconstruction,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 4501–4510, 2019

    work page 2019

  43. [43]

    Deephuman: 3d human reconstruction from a single image,

    Z. Zheng, T. Yu, Y . Wei, Q. Dai, and Y . Liu, “Deephuman: 3d human reconstruction from a single image,” inProceedings of the IEEE/CVF International Conference on Computer Vision, pp. 7739–7749, 2019

    work page 2019

  44. [44]

    M4esh: mmwave-based 3d human mesh construction for multiple subjects,

    H. Xue, Q. Cao, Y . Ju, H. Hu, H. Wang, A. Zhang, and L. Su, “M4esh: mmwave-based 3d human mesh construction for multiple subjects,” in Proceedings of the 20th ACM Conference on Embedded Networked Sensor Systems, pp. 391–406, 2022

    work page 2022

  45. [45]

    Towards generalized mmwave-based human pose estimation through signal augmentation,

    H. Xue, Q. Cao, C. Miao, Y . Ju, H. Hu, A. Zhang, and L. Su, “Towards generalized mmwave-based human pose estimation through signal augmentation,” inProceedings of the 29th Annual International Conference on Mobile Computing and Networking, pp. 1–15, 2023

    work page 2023

  46. [46]

    Texas instruments: Official website

    Texas Instruments, “Texas instruments: Official website.” https://www. ti.com/

    work page

  47. [47]

    Imagevk-74: Product page

    Mini-Circuits, “Imagevk-74: Product page.” https://www.minicircuits. com/WebStore/imagevk 74.html

    work page

  48. [48]

    Azure kinect dk: Product page

    Microsoft, “Azure kinect dk: Product page.” https://azure.microsoft.com/ en-us/products/kinect-dk

    work page

  49. [49]

    Optitrack motion capture: Official website

    OptiTrack, “Optitrack motion capture: Official website.” https:// optitrack.com/

    work page

  50. [50]

    Vicon help center / documentation wiki

    Vicon, “Vicon help center / documentation wiki.” https://vicon-help. atlassian.net/wiki/spaces/

    work page

  51. [51]

    Expressive body capture: 3d hands, face, and body from a single image,

    G. Pavlakos, V . Choutas, N. Ghorbani, T. Bolkart, A. A. Osman, D. Tzionas, and M. J. Black, “Expressive body capture: 3d hands, face, and body from a single image,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 10975– 10985, 2019

    work page 2019

  52. [52]

    Amass: Archive of motion capture as surface shapes,

    N. Mahmood, N. Ghorbani, N. F. Troje, G. Pons-Moll, and M. J. Black, “Amass: Archive of motion capture as surface shapes,” inProceedings of the IEEE/CVF international conference on computer vision, pp. 5442– 5451, 2019

    work page 2019

  53. [53]

    High-resolution frequency-wavenumber spectrum analysis,

    J. Capon, “High-resolution frequency-wavenumber spectrum analysis,” Proceedings of the IEEE, vol. 57, no. 8, pp. 1408–1418, 2005

    work page 2005

  54. [54]

    Os-cfar theory for multiple targets and nonuniform clutter,

    S. Blake, “Os-cfar theory for multiple targets and nonuniform clutter,” IEEE transactions on aerospace and electronic systems, vol. 24, no. 6, pp. 785–790, 1988

    work page 1988

  55. [55]

    On the continuity of rotation representations in neural networks,

    Y . Zhou, C. Barnes, J. Lu, J. Yang, and H. Li, “On the continuity of rotation representations in neural networks,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 5745–5753, 2019

    work page 2019