Pith. sign in

REVIEW 1 major objections 4 minor 52 references

RETR: Multi-View Radar Detection Transformer for Indoor Perception

T0 review · 1 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Extending DETR to two radar heatmap views, with depth-prioritized attention, a tri-plane loss, and a learnable radar-to-camera transform, yields the best reported indoor radar detection and segmentation: 15.38+ AP and 11.91+ IoU over…

desk verdict Solid DETR-for-radar engineering, but the advertised 15.38 AP SOTA margin is measured against a modified baseline on refined labels, so the headline claim needs verification. read the letter →

arxiv 2411.10293 v3 pith:R62VQBID submitted 2024-11-15 cs.CV cs.AIcs.LGmath.DG

Ryoma Yataka , Adriano Cardace , Pu Perry Wang , Petros Boufounos , Ryuhei Takahashi This is my paper · ORCID
classification cs.CVcs.AIcs.LGmath.DG
keywords multi-viewradarperceptiondetectiontransformerDETRextensionindoorinstancesegmentationtunablepositionalencodingradar-to-cameracalibration
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

RETR is a transformer-based detector that takes two radar heatmaps, one horizontal (azimuth-depth) and one vertical (elevation-depth), and produces image-plane bounding boxes and instance masks. The paper's claim is that by treating the two views as sharing a depth axis and letting a tunable positional encoding prioritize depth similarity, the DETR-style set-prediction machinery can associate the views without hand-crafted proposals or non-maximum suppression. The authors report that RETR beats prior methods by 15.38+ average precision for detection and 11.91+ IoU for segmentation on two indoor radar datasets. A tri-plane loss supervises the predicted 3D boxes in both radar planes and the image plane, and a learnable SO(3) transform replaces a calibrated radar-to-camera mapping. If true, this would make radar a practical, privacy-preserving sensor for indoor monitoring tasks that normally require cameras.

What carries the argument

The load-bearing object is the tunable positional encoding (TPE), a positional embedding split into a depth part and an angular part whose dimensions are controlled by a single ratio alpha. Because the horizontal and vertical heatmaps share the depth axis, dot products of depth embeddings stay meaningful across views while angular embeddings compare azimuth with elevation. In RETR's self-attention and cross-attention, TPE makes queries and keys at similar depths score higher, and the paper adds a tri-plane loss and a learnable SO(3) transform on top of this mechanism. The encoder pools top-K strongest features from each view and lets self-attention associate them, while the decoder uses object queries with 3D spatial embeddings to attend to both views.

What would settle it

Evaluate RETR on a two-view radar capture with deliberately shifted range bins between the views or with a non-co-located array pair; a small accuracy drop compared with the aligned setting would show the shared-depth assumption is not load-bearing, while a large drop would confirm it. Separately, measuring AP on an array mount with known non-rigid distortion would test whether the rigid SO(3) transform is essential.

Watch

Extended reading notes

Core claim

The central claim is that multi-view radar perception should not be treated as an image problem; the distinctive geometry of two radar views, especially the shared depth axis, should be built into the attention architecture. RETR does this in three moves: a depth-prioritized positional encoding (TPE) that makes cross-view attention prefer tokens at similar depths, a tri-plane set-prediction loss that evaluates the same 3D box in the horizontal radar plane, vertical radar plane, and image plane, and a learnable radar-to-camera transformation parameterized through the exponential map so the rotation stays on SO(3). The paper reports that on the MMVR P2S1 protocol RETR reaches 46.75 AP versus 31.37 for RFMask, a 15.38-point gain, and 77.21 IoU versus 65.30, an 11.91-point gain; on HIBER WALK the detection gain is 4.32 AP.

Load-bearing premise

The method assumes the horizontal and vertical radar heatmaps share one aligned depth axis and that the radar-to-camera mapping is a single rigid rotation plus translation; if the views are not synchronized or the geometry is non-rigid, the depth-prioritized attention and the projection supervision lose their justification.

Editorial extensions

If this is right

  • End-to-end radar detection and segmentation no longer need proposal networks or non-maximum suppression, matching DETR's pipeline design for images.
  • Vertical-view supervision is load-bearing: adding the vertical radar plane to the loss improves AP by 4.47 and IoU by 2.06 on the main benchmark.
  • The learnable radar-to-camera transform can replace a separate calibration step, improving AP from 42.97 to 46.75 on the MMVR P2S1 protocol.
  • The performance gain over the proposal-based baseline is much larger for diverse activities such as sitting and writing than for walking-only scenes, because fixed-height proposals fail when people are not standing.
  • RETR runs at about 42 frames per second on an A40 GPU, so the accuracy gain does not come at the cost of real-time operation.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The TPE idea, a shared-axis depth-prioritizing positional encoding, is not radar-specific; any two-view sensing setup with a common range coordinate, such as stereo cameras or LiDAR-camera pairs, could adopt it, so the architectural lesson may transfer.
  • The reported margins depend on two indoor radar datasets with co-located arrays; an untested but natural next step is evaluating RETR on automotive 4D radar heatmaps, where the shared-depth assumption also holds but clutter and multipath are different.
  • Because the authors refine the HIBER ground-truth boxes, some of the gap over the baseline may reflect label quality as well as architecture; re-benchmarking with the original labels would isolate the contribution of the model design.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

1 major / 4 minor

Summary. This paper presents RETR, a DETR-style transformer for object detection and instance segmentation from two-view indoor radar heatmaps (horizontal azimuth-depth and vertical elevation-depth). RETR replaces hand-crafted proposals and NMS with transformer encoder-decoder associations and contributes a tunable positional encoding (TPE) that changes the dimension split between depth and angle embeddings, a tri-plane set-prediction loss that supervises 3D radar boxes in the horizontal radar, vertical radar, and image planes, and a learnable radar-to-camera rigid transformation parameterized via the SO(3) exponential map. The method is evaluated on the MMVR and HIBER datasets, with ablations on the TPE ratio, top-K size, tri-plane loss, learnable transformation, and training data size, plus instance segmentation results on MMVR.

Significance. The method is a clean adaptation of DETR to multi-view radar heatmaps, and its derivations are technically sound: the TPE factorization in Eq. (12) is an algebraic identity, the SO(3) reparameterization in Eq. (18) is standard, and the complexity analysis in Appendix A is coherent. The code is publicly released. If the reported margins over the actual published baselines hold, RETR would be a strong new state of the art for indoor radar image-plane perception, and the TPE and tri-plane ideas are likely transferable to other multi-view range-sensor settings. The main caveat is that the abstract's 'state-of-the-art' comparison is not demonstrated against the original published methods, because the RFMask and DETR baselines were modified and the HIBER labels were refined. This does not invalidate the internal ablation story, but it materially changes the strength of the headline claim.

major comments (1)
  1. [Abstract; Section 5.1; Appendix E; Tables 1 and 4] The headline claim that RETR 'outperforms existing state-of-the-art methods by a margin of 15.38+ AP for object detection and 11.91+ IoU for instance segmentation' is not supported by the comparison actually reported. Table 1 compares RETR (46.75 AP) with an RFMask variant (31.37 AP) that, according to Section 5.1, was enhanced with a unified bi-plane H+I loss, and Table 4 compares RETR with the same modified RFMask. Appendix E further states that the HIBER BBox labels were refined and that RFMask was modified by adding an image-plane BBox regression module because the original RFMask cannot predict refined BBoxes. The paper does not report the original published RFMask numbers, nor a run of the unmodified RFMask on the original or refined labels. Without those numbers, the 15.38 AP and 11.91 IoU margins could reflect the baseline modifications or label refinement rather than the RETR contributions. Please add the original-baseline comparison (or a careful decomposition) and, ideally, repeated-seed statistics, or rephrase the abstract to state that the comparison is against re-implemented, modified baselines.
minor comments (4)
  1. [Section 5.3, Table 3a] The text says Table 3a reports AP50 and AP75, but the table lists only AP and AR1; the AP50/AP75 behavior appears only in Figure 13 of Appendix G. Please align the table and text.
  2. [Appendix B, Table 4] The text says RETR achieves 77.07 IoU, but Table 4 reports 77.21; the abstract's 11.91 margin also corresponds to 77.21, so the in-text number should be corrected.
  3. [Appendix G, before Table 6] The sentence reporting 'the improvement is 15.28 AP from DETR to RETR' does not match Table 6, where the difference is 28.98 - 17.00 = 11.98 AP; please correct this and clarify that the surrounding discussion refers to the MULTI split, not WALK.
  4. [Eq. (18)] There are typos: 's.t. [ω] = ...' and 'With the above reparameterization, the learnable radar-to-camera coordinate transformation in Eq. 7 reduces' has an ungrammatical capital W; also the 'approx' symbol should be an equality because Eq. (18) is the exact Rodrigues formula for the exponential map on SO(3).

Circularity Check

0 steps flagged · score 2.0 of 10

No derivation-level circularity; the claimed gains are measured against modified baselines on a self-created benchmark, but no prediction reduces to a fitted value by the paper's own equations.

full rationale

The paper's derivation chain is self-contained at the equation level. TPE (Eqs. 10-13) is an algebraic decomposition of the conditional-DETR concatenation operation; the dimension-splitting ratio alpha is a hyperparameter tuned on MMVR validation, not a fitted quantity that is later renamed as a prediction. The learnable radar-to-camera transformation (Eq. 18) is the standard SO(3) exponential-map reparameterization, and the tri-plane loss (Eq. 17) is a sum of projection losses, neither of which smuggles the target result into its inputs. The 3D-to-2D projection and set-prediction losses are conventional DETR machinery adapted to radar coordinates. The main caveat is empirical rather than circular: the headline 15.38 AP / 11.91 IoU margins in Tables 1 and 4 are computed against RFMask and DETR variants that were explicitly modified (unified H+I loss, top-K selection, added image-plane BBox regression) and against refined HIBER labels, as stated in Section 5.1 and Appendix E. This means the advertised comparison to 'existing state-of-the-art methods' is not verified against the original published implementations, but this is a benchmark-comparison and correctness concern, not a case where an equation reduces to its own input. Similarly, MMVR [26] is a same-group benchmark and alpha=0.6 and K=256 are tuned on it, but the method is also evaluated on the external HIBER dataset and shows consistent, if smaller, gains. Because no prediction is forced by construction and no load-bearing argument reduces to a self-citation, the circularity score is low; the minor self-citation and benchmark-tuning caveats justify a score of 2 rather than 0.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

RETR introduces no new physical entities or conserved quantities; all assumptions are domain models for radar and camera geometry. The free parameters are standard architecture and loss hyperparameters, with alpha and K tuned on the target benchmarks.

free parameters (4)
  • TPE depth dimension ratio alpha = 0.6
    Controls the split of the positional embedding dimension between depth and angle; tuned on MMVR validation (ablation in Table 3a) and reused for HIBER.
  • Top-K selection count K = 256
    Number of strongest radar features kept from each view; chosen from ablation (Table 10) where K=256 gave the best AP.
  • Loss weights lambda_GIoU and lambda_L1 per plane = varies per dataset (Table 5)
    Weighting of GIoU and L1 losses on horizontal, vertical, and image planes; set by hand per dataset.
  • Number of input frames T = 4
    Temporal context length; chosen as part of the model configuration.
assumptions (5)
  • domain assumption The radar heatmap is modeled as a sum of point scatterers with round-trip distances (Eq. 1).
    This linear scattering model underlies the generation of horizontal and vertical heatmaps; real radar returns include multipath and noise.
  • domain assumption Horizontal and vertical radar views share a common depth axis with aligned range bins (Section 3).
    The entire TPE design and cross-view association rely on depth being shared and consistent between the two views.
  • domain assumption Radar-to-camera transformation is a rigid transform: g_camera = R*g_radar + t with R in SO(3) (Eq. 7).
    Used to project 3D radar boxes to the image plane; if the sensor setup has non-rigid distortion or unknown scaling, this parameterization is insufficient.
  • standard math Pinhole camera model for 3D-to-2D projection (Section 3).
    Standard assumption for camera projection.
  • standard math Hungarian matching and DETR set-prediction loss formulation (Section 4.4).
    Adopted from DETR; not a new claim.

how reviews work

0 comments
Cite this review

Pith. "Pith review of RETR: Multi-View Radar Detection Transformer for Indoor Perception." pith.science (2026). https://pith.science/paper/R62VQBID

@misc{pith2026241110293,
  author       = {Pith},
  title        = {Pith review of: RETR: Multi-View Radar Detection Transformer for Indoor Perception},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/R62VQBID}},
  note         = {Machine review of arXiv:2411.10293}
}
read the original abstract

Indoor radar perception has seen rising interest due to affordable costs driven by emerging automotive imaging radar developments and the benefits of reduced privacy concerns and reliability under hazardous conditions (e.g., fire and smoke). However, existing radar perception pipelines fail to account for distinctive characteristics of the multi-view radar setting. In this paper, we propose Radar dEtection TRansformer (RETR), an extension of the popular DETR architecture, tailored for multi-view radar perception. RETR inherits the advantages of DETR, eliminating the need for hand-crafted components for object detection and segmentation in the image plane. More importantly, RETR incorporates carefully designed modifications such as 1) depth-prioritized feature similarity via a tunable positional encoding (TPE); 2) a tri-plane loss from both radar and camera coordinates; and 3) a learnable radar-to-camera transformation via reparameterization, to account for the unique multi-view radar setting. Evaluated on two indoor radar perception datasets, our approach outperforms existing state-of-the-art methods by a margin of 15.38+ AP for object detection and 11.91+ IoU for instance segmentation, respectively. Our implementation is available at https://github.com/merlresearch/radar-detection-transformer.

Figures

Figures reproduced from arXiv: 2411.10293 by the authors.

Figure 1
Figure 1. By taking horizontal-view and vertical-view radar heatmaps as inputs, RETR introduces a [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Indoor radar perception pipeline: (a) multi-radar views are utilized to estimate 3D BBoxes [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The RETR architecture: 1) Encoder: Top-K features selection and tunable positional encoding to assist feature association across the two radar views; 2) Decoder: TPE is also used to assist the association between object queries and multi-view radar features; 3) 3D BBox Head: Object queries are enforced to estimate 3D objects in the radar coordinate and projected to 3 planes for supervision via a coordinate transform… view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Schemes of positional encoding: (a) the sum operation in the original DETR; (b) the [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Tri-Plane BBox loss. Since RETR predicts 3D BBoxes g¯ in the 3D radar coordinate and maps them into the 2D image plane, we propose to enhance the above Hungarian match cost matrix using a Tri-Plane BBox Loss from both the radar coordinate and image plane. This is illus…
Figure 6
Figure 6. Figure 6: Visualization of cross-attention map between predicted BBoxes and multi-view radar [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Illustration of (left) encoder and (right) decoder of RETR. [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]
Figure 8
Figure 8. Figure 8: Illustration of segmentation head. d is the embedding dimension and K is the number of selected features from the Top-K selection. The cost of computing a single query/key/value embedding is O(d ′d) (with d = M d′ where M denotes the number of attention heads and d ′ t…
Figure 9
Figure 9. Figure 9: Visualization of segmentation results. the latter strategy. During prediction, we filter out the detection with a confidence below 50%, then compute the per-pixel argmax to produce the final binary segmentation mask. Main Results: We report quantitative results the seg…
Figure 10
Figure 10. Figure 10: Visualization of TPE: (top row) positional embeddings for each axis; depth and angle, [PITH_FULL_IMAGE:figures/full_fig_p018_10.png]
Figure 11
Figure 11. Figure 11: Multi-view heatmap preprocessing [PITH_FULL_IMAGE:figures/full_fig_p018_11.png]
Figure 12
Figure 12. Figure 12: Original (left) versus Refined (right) BBoxes in the HIBER dataset. HIBER Dataset: HIBER [38] is an open-source multi-view radar dataset including horizontal and vertical radar heatmaps and annotations such as 2D and 3D poses, BBoxes, and segmentation masks. Among its…
Figure 13
Figure 13. Figure 13: Full figure: Impact of tunable dimen￾sion ratio α in TPE. Impact of Tunable Dimension Ratio α in TPE: To investigate the impact of tuning in TPE, we observed the performance differences when vary￾ing the ratio of depth and angle dimensions (since the total dimension i…
Figure 14
Figure 14. Figure 14: Visualization and comparison between RETR and RFMask. Each row indicates the [PITH_FULL_IMAGE:figures/full_fig_p023_14.png]
Figure 15
Figure 15. Figure 15: Visualization of failure cases. Each row indicates the segment name used from the “P2S1” [PITH_FULL_IMAGE:figures/full_fig_p024_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 25 canonical work pages

  1. [1]

    Capturing the human figure through a wall

    Fadel Adib, Chen-Yu Hsu, Hongzi Mao, Dina Katabi, and Frédo Durand. Capturing the human figure through a wall. ACM Trans. Graph., 34(6), 2015. URL https://doi.org/10.1145/ 2816795.2818072

    arXiv 2015

  2. [2]

    mRI: Multi-modal 3D human pose esti- mation dataset using mmWave, RGB-D, and inertial sensors

    Sizhe An, Yin Li, and Umit Ogras. mRI: Multi-modal 3D human pose esti- mation dataset using mmWave, RGB-D, and inertial sensors. In Advances in Neural Information Processing Systems , volume 35, pp. 27414–27426, 2022. URL https://proceedings.neurips.cc/paper_files/paper/2022/file/ af9c9c6d2da701da5a0acf91ec217815-Paper-Datasets_and_Benchmarks.pdf

    work page 2022

  3. [3]

    End-to-end object detection with transformers

    Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, and Sergey Zagoruyko. End-to-end object detection with transformers. In European Confer- ence on Computer Vision (ECCV), pp. 213–229, 2020. URL https://doi.org/10.1007/ 978-3-030-58452-8_13

    work page 2020

  4. [5]

    RadarOcc: Robust 3D occupancy prediction with 4D imaging radar

    Fangqiang Ding, Xiangyu Wen, Yunzhou Zhu, Yiming Li, and Chris Xiaoxuan Lu. RadarOcc: Robust 3D occupancy prediction with 4D imaging radar. arXiv:2405.14014, 2024. URL https://arxiv.org/abs/2405.14014

    arXiv 2024

  5. [6]

    Mark Everingham, Luc Van Gool, Christopher K. I. Williams, John Winn, and Andrew Zisser- man. The PASCAL Visual Object Classes (VOC) Challenge.International Journal of Computer Vision, 88(2):303–338, 2010. URL https://doi.org/10.1007/s11263-009-0275-4

    doi:10.1007/s11263-009-0275-4 2010

  6. [7]

    Sevgi Zubeyde Gurbuz and Moeness G. Amin. Radar-based human-motion recognition with deep learning: Promising applications for indoor monitoring. IEEE Signal Processing Magazine, 36(4):16–28, 2019. URL https://doi.org/10.1109/MSP.2018.2890128

    arXiv 2019

  7. [8]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 770–778, 2016. URL https://doi.org/10.1109/CVPR.2016.90

    doi:10.1109/cvpr.2016.90 2016

  8. [9]

    What makes for effective detection proposals? IEEE Transactions on Pattern Analysis and Machine Intelligence, 38(4): 814–830, 2016

    Jan Hosang, Rodrigo Benenson, Piotr Dollár, and Bernt Schiele. What makes for effective detection proposals? IEEE Transactions on Pattern Analysis and Machine Intelligence, 38(4): 814–830, 2016. URL https://doi.org/10.1109/TPAMI.2015.2465908

    work page arXiv 2016

Show all 52 references
  1. [10]

    DAC-DETR: Divide the attention layers and conquer

    Zhengdong Hu, Yifan Sun, Jingdong Wang, and Yi Yang. DAC-DETR: Divide the attention layers and conquer. In Advances in Neural Information Processing Systems , volume 36, pp. 75189–75200, 2023. URL https://proceedings.neurips.cc/paper_files/paper/ 2023/file/edd0d433f8a1a51aa112...

    2023

  2. [11]

    RTNH+: Enhanced 4D radar object detection network using two-level preprocessing and vertical encoding

    Seung-Hyun Kong, Dong-Hee Paek, and Sangyeong Lee. RTNH+: Enhanced 4D radar object detection network using two-level preprocessing and vertical encoding. IEEE Transactions on Intelligent Vehicles, pp. 1–14, 2024. URL https://10.1109/TIV.2024.3428696

    2024

  3. [12]

    Harold W. Kuhn. The Hungarian method for the assignment problem. Naval research logistics quarterly, 2(1-2):83–97, 1955. URL https://doi.org/10.1007/978-3-540-68279-0_2

    1955 doi

  4. [13]

    John. M. Lee. Introduction to Smooth Manifolds. Springer, 2003. URL https://doi.org/ 10.1007/978-1-4419-9982-5

    2003 doi

  5. [14]

    HuPR: A benchmark for human pose estimation using millimeter wave radar

    Shih-Po Lee, Niraj Prakash Kini, Wen-Hsiao Peng, Ching-Wen Ma, and Jenq-Neng Hwang. HuPR: A benchmark for human pose estimation using millimeter wave radar. In IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pp. 5715–5724, 2023. URL https://doi.org/10.110...

    2023

  6. [15]

    Exploiting temporal relations on radar perception for autonomous driving

    Peizhao Li, Pu Wang, Karl Berntorp, and Hongfu Liu. Exploiting temporal relations on radar perception for autonomous driving. In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 17050–17059, 2022. URL https://openaccess.thecvf. com/content/CVPR2022/pa...

    2022

  7. [16]

    Focal loss for dense object detection

    Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Dollár. Focal loss for dense object detection. In IEEE/CVF International Conference on Computer Vision (ICCV) , pp. 2999–3007, 2017. URL https://doi.org/10.1109/ICCV.2017.324

    2017 doi

  8. [17]

    DAB-DETR: Dynamic anchor boxes are better queries for DETR

    Shilong Liu, Feng Li, Hao Zhang, Xiao Yang, Xianbiao Qi, Hang Su, Jun Zhu, and Lei Zhang. DAB-DETR: Dynamic anchor boxes are better queries for DETR. In International Conference on Learning Representations (ICLR), 2022. URL https://openreview.net/forum?id= oMI9PjOb9Jl

    2022

  9. [18]

    Detection transformer with stable matching

    Shilong Liu, Tianhe Ren, Jiayu Chen, Zhaoyang Zeng, Hao Zhang, Feng Li, Hongyang Li, Jun Huang, Hang Su, Jun Zhu, and Lei Zhang. Detection transformer with stable matching. In IEEE/CVF International Conference on Computer Vision (ICCV), pp. 6468–6477, 2023. URL https://doi.org...

    2023

  10. [19]

    Echoes beyond points: Unleashing the power of raw radar data in multi-modality fusion

    Yang Liu, Feng Wang, Naiyan Wang, and Zhao-Xiang Zhang. Echoes beyond points: Unleashing the power of raw radar data in multi-modality fusion. In Advances in Neural Information Processing Systems , volume 36, pp. 53964–53982,

  11. [20]

    Saputra, Peijun Zhao, Yasin Almalioglu, Pedro P

    Chris Xiaoxuan Lu, Muhamad Risqi U. Saputra, Peijun Zhao, Yasin Almalioglu, Pedro P. B. de Gusmao, Changhao Chen, Ke Sun, Niki Trigoni, and Andrew Markham. milliEgo: single- chip mmwave radar aided egomotion estimation via deep sensor fusion. In The 18th Conference on Embedded...

    2020

  12. [22]

    V-Net: Fully convolutional neural networks for volumetric medical image segmentation

    Fausto Milletari, Nassir Navab, and Seyed-Ahmad Ahmadi. V-Net: Fully convolutional neural networks for volumetric medical image segmentation. In International Conference on 3D Vision (3DV), pp. 565–571, 2016. URL https://10.1109/3DV.2016.79

    2016 doi

  13. [23]

    Multi- view radar semantic segmentation

    Arthur Ouaknine, Alasdair Newson, Patrick Pérez, Florence Tupin, and Julien Rebut. Multi- view radar semantic segmentation. In IEEE/CVF International Conference on Computer Vi- sion (ICCV), pp. 15651–15660, 2021. URL https://doi.org/10.1109/ICCV48922.2021. 01538

    2021

  14. [24]

    K-Radar: 4D radar object detection for autonomous driving in various weather conditions

    Dong-Hee Paek, Seung-Hyun Kong, and Kevin Tirta Wijaya. K-Radar: 4D radar object detection for autonomous driving in various weather conditions. In Advances in Neural Information Processing Systems , volume 35, pp. 3819–3829,

  15. [25]

    Rank-DETR for high quality object detection

    Yifan Pu, Weicong Liang, Yiduo Hao, Yuhui Yuan, Yukang Yang, Chao Zhang, Han Hu, and Gao Huang. Rank-DETR for high quality object detection. In Advances in Neural Information Processing Systems , volume 36, pp. 16100–16113,

  16. [26]

    Mahbubur Rahman, Ryoma Yataka, Sorachi Kato, Pu Wang, Peizhao Li, Adriano Cardace, and Petros Boufounos

    M. Mahbubur Rahman, Ryoma Yataka, Sorachi Kato, Pu Wang, Peizhao Li, Adriano Cardace, and Petros Boufounos. MMVR: Millimeter-wave multi-view radar dataset and benchmark for indoor perception. In European Conference on Computer Vision (ECCV), pp. 306–322, 2025. ISBN 978-3-031-7...

    2025 doi

  17. [27]

    Faster R-CNN: Towards real-time object detection with region proposal networks

    Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster R-CNN: Towards real-time object detection with region proposal networks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 39(6):1137–1149, 2017. URL https://doi.org/10.1109/TPAMI. 2016.2577031

    2017

  18. [28]

    URL https://proceedings.neurips.cc/paper_files/paper/2023/file/ 34074479ee2186a9f236b8fd03635372-Paper-Conference.pdf

    2023

  19. [29]

    U-Net: Convolutional net- works for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-Net: Convolutional net- works for biomedical image segmentation. In Medical Image Computing and Computer- Assisted Intervention (MICCAI) , pp. 234–241, 2015. URL https://doi.org/10.1007/ 978-3-319-24574-4_28

    2015

  20. [30]

    mm-Pose: Real-time human skeletal posture estimation using mmWave radars and CNNs

    Arindam Sengupta, Feng Jin, Renyuan Zhang, and Siyang Cao. mm-Pose: Real-time human skeletal posture estimation using mmWave radars and CNNs. IEEE Sensors Journal, 20(17): 10032–10044, 2020. URL https://doi.org/10.1109/JSEN.2020.2991741. 12

    2020

  21. [31]

    Generalized intersection over union: A metric and a loss for bounding box regression

    Hamid Rezatofighi, Nathan Tsoi, JunYoung Gwak, Amir Sadeghian, Ian Reid, and Silvio Savarese. Generalized intersection over union: A metric and a loss for bounding box regression. In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 658–666,

  22. [32]

    Human detection from 4D radar data in low-visibility field conditions

    Mikael Skog, Oleksandr Kotlyar, Vladimír Kubelka, and Martin Magnusson. Human detection from 4D radar data in low-visibility field conditions. arXiv:2404.05307, 2024. URL https: //arxiv.org/abs/2404.05307

    2024 arXiv

  23. [33]

    A micro Lie theory for state estimation in robotics

    Joan Solá, Jeremie Deray, and Dinesh Atchuthan. A micro Lie theory for state estimation in robotics. arXiv:1812.01537, 2021. URL https://arxiv.org/abs/1812.01537

    2021 arXiv

  24. [34]

    Petropulu, and H

    Shunqiao Sun, Athina P. Petropulu, and H. Vincent Poor. MIMO radar for advanced driver- assistance systems and autonomous driving: Advantages and challenges. IEEE Signal Process- ing Magazine, 37(4):98–117, 2020. URL https://10.1109/MSP.2020.2978507

    2020

  25. [35]

    RadHAR: Human activity recognition from point clouds generated through a millimeter-wave radar

    Akash Deep Singh, Sandeep Singh Sandha, Luis Garcia, and Mani Srivastava. RadHAR: Human activity recognition from point clouds generated through a millimeter-wave radar. In The 3rd ACM Workshop on Millimeter-Wave Networks and Sensing Systems, mmNets ’19, pp. 51–56,

  26. [36]

    URL https://doi.org/10.1145/3349624.3356768

  27. [37]

    Anchor DETR: Query design for transformer-based detector

    Yingming Wang, Xiangyu Zhang, Tong Yang, and Jian Sun. Anchor DETR: Query design for transformer-based detector. AAAI Conference on Artificial Intelligence, 36(3):2567–2575, 2022. URL https://doi.org/10.1609/aaai.v36i3.20158

    2022 doi

  28. [38]

    RFMask: A simple baseline for human silhouette segmentation with radio signals

    Zhi Wu, Dongheng Zhang, Chunyang Xie, Cong Yu, Jinbo Chen, Yang Hu, and Yan Chen. RFMask: A simple baseline for human silhouette segmentation with radio signals. IEEE Transactions on Multimedia, 25:4730–4741, 2023. URL https://doi.org/10.1109/TMM. 2022.3181455

    2023

  29. [39]

    mmMesh: towards 3D real-time dynamic human mesh construction using millimeter-wave

    Hongfei Xue, Yan Ju, Chenglin Miao, Yijiang Wang, Shiyang Wang, Aidong Zhang, and Lu Su. mmMesh: towards 3D real-time dynamic human mesh construction using millimeter-wave. In MobiSys, pp. 269–282, 2021. URL https://doi.org/10.1145/3458864.3467679

    2021

  30. [40]

    Indoor person identification using a low-power FMCW radar

    Baptist Vandersmissen, Nicolas Knudde, Azarakhsh Jalalvand, Ivo Couckuyt, André Bourdoux, Wesley De Neve, and Tom Dhaene. Indoor person identification using a low-power FMCW radar. IEEE Transactions on Geoscience and Remote Sensing, 56(7):3941–3952, 2018. URL https://doi.org/1...

    2018

  31. [41]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. In Advances in Neural Infor- mation Processing Systems, volume 30, 2017. URL https://proceedings.neurips.cc/ paper_files/pape...

    2017

  32. [42]

    Jaakkola, and Matt T

    Mingmin Zhao, Shichao Yue, Dina Katabi, Tommi S. Jaakkola, and Matt T. Bianchi. Learn- ing sleep stages from radio signals: A conditional adversarial architecture. In Interna- tional Conference on Machine Learning (ICML) , volume 70, pp. 4100–4109, 2017. URL https://proceeding...

    2017

  33. [43]

    Through-wall human pose estimation using radio signals

    Mingmin Zhao, Tianhong Li, Mohammad Abu Alsheikh, Yonglong Tian, Hang Zhao, Antonio Torralba, and Dina Katabi. Through-wall human pose estimation using radio signals. In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 7356–7365,

  34. [44]

    RF-based 3D skeletons

    Mingmin Zhao, Yonglong Tian, Hang Zhao, Mohammad Abu Alsheikh, Tianhong Li, Rumen Hristov, Zachary Kabelac, Dina Katabi, and Antonio Torralba. RF-based 3D skeletons. In The 2018 Conference of the ACM Special Interest Group on Data Communication (SIGCOMM), pp. 267–281, 2018. UR...

    2018

  35. [45]

    MM-Fi: Multi- modal non-intrusive 4D human dataset for versatile wireless sensing

    Jianfei Yang, He Huang, Yunjiao Zhou, Xinyan Chen, Yuecong Xu, Sheng- hai Yuan, Han Zou, Chris Xiaoxuan Lu, and Lihua Xie. MM-Fi: Multi- modal non-intrusive 4D human dataset for versatile wireless sensing. In Ad- vances in Neural Information Processing Systems , volume 36, pp....

  36. [46]

    URL https://proceedings.neurips.cc/paper_files/paper/2023/file/ 3baf7a39d07e9f4f1e258a412df94521-Paper-Datasets_and_Benchmarks.pdf

    2023

  37. [47]

    SIRA: Scalable inter-frame relation and association for radar perception

    Ryoma Yataka, Pu Wang, Petros Boufounos, and Ryuhei Takahashi. SIRA: Scalable inter-frame relation and association for radar perception. In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 15024–15034, 2024. doi: 10.1109/CVPR52733.2024.01423. URL http...

    2024

  38. [52]

    Human tracking and identification through a millimeter wave radar

    Peijun Zhao, Chris Xiaoxuan Lu, Jianan Wang, Changhao Chen, Wei Wang, Niki Trigoni, and Andrew Markham. Human tracking and identification through a millimeter wave radar. Ad Hoc Networks, 116, 2021. URL https://doi.org/10.1016/j.adhoc.2021.102475

    2021

  39. [53]

    Cubelearn: End-to-end learning for human motion recognition from raw mmWave radar signals

    Peijun Zhao, Chris Xiaoxuan Lu, Bing Wang, Niki Trigoni, and Andrew Markham. Cubelearn: End-to-end learning for human motion recognition from raw mmWave radar signals. IEEE Internet of Things Journal, 10(12):10236–10249, 2023. URL https://doi.org/10.1109/ JIOT.2023.3237494

    2023

  40. [54]

    P1” is used to establish the best possible radar perception benchmarks, while “P2

    Xizhou Zhu, Weijie Su, Lewei Lu, Bin Li, Xiaogang Wang, and Jifeng Dai. Deformable DETR: Deformable transformers for end-to-end object detection. In International Conference on Learning Representations (ICLR), 2021. URL https://openreview.net/forum?id= gZ9hCDWe6ke. 14 A Detail...

    2021

  41. [2018]

    URL https://doi.org/10.1109/CVPR.2018.00768. 13

    2018

  42. [2019]

    URL https://doi.org/10.1109/CVPR.2019.00075

    2019

  43. [2022]

    URL https://proceedings.neurips.cc/paper_files/paper/2022/file/ 185fdf627eaae2abab36205dcd19b817-Paper-Datasets_and_Benchmarks.pdf

    2022

  44. [2023]

    URL https://proceedings.neurips.cc/paper_files/paper/2023/file/ a8f7f12b29d9b8c227785f6b529f63b7-Paper-Conference.pdf

    2023

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.