Pith. sign in

REVIEW 4 major objections 4 minor 1 cited by

Coop-WD: Cooperative Perception with Weighting and Denoising for Robust V2V Communication

T0 review · 4 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A two-stage receiver fix—weighting then denoising—keeps cooperative 3D detection accurate on impaired V2V channels.

desk verdict A useful engineering combo of self-supervised weighting and diffusion denoising for V2V-robust cooperative perception, but the paper's central claim of consistent superiority is contradicted by its own Table I. read the letter →

arxiv 2505.03528 v2 pith:HOOHV2JJ submitted 2025-05-06 cs.CV

Chenguang Liu , Jianjun Chen , Yunfei Chen , Yubei He , Zhuangkun Wei , Hongjian Sun , Haiyan Lu , Qi Hao This is my paper · ORCID
classification cs.CV
keywords cooperativeperceptionV2Vcommunication3Dobjectdetectionconditionaldiffusionmodelself-supervisedweightingnon-stationarychannelfeaturedenoisingintermediatefusion
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

Coop-WD is a receiver-side enhancement for cooperative perception that treats two failure scales separately: a vehicle-level scalar weighting suppresses whole corrupted contributions, and a pixel-level conditional diffusion model reconstructs fine-grained feature distortion. The paper aims to show that this joint scheme keeps 3D detection accuracy high across a range of V2V impairments, from severe to mild, including time-varying distortion and imperfect channel state information, rather than only at one noise level. If true, it would make shared perception usable under realistic radio conditions, which matters because a single vehicle's sensors are limited by occlusion and range while V2V sharing is limited by channel reliability. The paper also proposes an economical variant that disables the expensive denoiser when the weighting signal indicates severe distortion, preserving accuracy while cutting compute.

What carries the argument

The carrying mechanism is a conditional diffusion probabilistic model operating on shared feature maps. At diffusion step $t$, the corrupted feature $x_t$ is drawn from $q_{\text{cond}}(x_t|x_0,y)=\mathcal{N}(x_t;(1-m_t)\sqrt{\bar\alpha_t}x_0+m_t\sqrt{\bar\alpha_t}y,\delta_t I)$, a Gaussian blend of the clean transmitted feature $x_0$ and the received distorted feature $y$, with mixing ratio $m_t$ growing from 0 to about 1. The reverse process, a U-Net that predicts the mixed noise $\epsilon_\theta(x_t,y,t)$, iteratively removes that blend to recover the denoised feature map. Above it sits a self-supervised contrastive weighting module that multiplies each received feature by a scalar $W_k\in[0,1]$; its output feeds the diffusion model. The weighting handles severe whole-vehicle corruption, the diffusion handles residual pixel-level distortion, and the two modules share residual blocks and skip connections so they can be trained and run together.

What would settle it

Record real V2V feature corruption that includes packet bursts and correlated interference, feed those traces through the same trained pipeline, and compare Coop-WD against Coop-W; the central claim predicts the diffusion stage still lifts AP, so observing Coop-WD at or below Coop-W under such non-Gaussian corruption would falsify it.

Watch

Extended reading notes

Core claim

The paper argues that combining a CAV-level weighting module with a pixel-level conditional diffusion denoiser yields a cooperative-perception system that outperforms both components alone and the conventional CNN-autoencoder baseline across every channel condition tested: simulated Rician fading, realistic WINNER II, and a non-stationary V2V model, at SNRs from 0 to 20 dB. The ablation results show Coop-WD consistently beats Coop-W and Coop-D at IoU 0.5 and 0.7; in the non-stationary V2V channel at 0 dB, for example, it reaches AP 0.487 at IoU 0.5 while Coop-W reaches 0.483 and Coop-D reaches 0.151. The same hierarchy transfers to an attentive fusion backend, and the eco variant cuts runtime by more than half under severe distortion with near-equal AP, while saving roughly 10% runtime at higher SNR.

Load-bearing premise

The whole denoising gain rests on treating the damage a V2V channel does to a feature map as Gaussian noise blended with the clean signal; if real corruption is bursty, correlated, or state-dependent in a way the training distribution never covers, the reverse process may not reconstruct the clean features and the advantage over weighting alone could disappear.

Editorial extensions

If this is right

  • Coop-WD consistently beats Coop-W, Coop-D, and the CNN-autoencoder baseline at every tested SNR on Rician, WINNER II, and non-stationary V2V channels.
  • Ablation shows neither module alone is sufficient: weighting rescues severe distortion while denoising handles mild distortion, and their combination dominates both.
  • The framework transfers to an attentive fusion backend, not only V2VNet, so the benefit generalizes across fusion designs within the tested settings.
  • Training on a non-stationary V2V channel and testing on unseen Rician and WINNER II channels still yields the same ordering, indicating cross-channel generalization.
  • Coop-WD-eco cuts runtime by more than half under severe distortion with minimal AP loss and saves about 10% at higher SNR, making the added compute controllable.

Reading between the lines

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

  • The corruption model in Eq. (7) is Gaussian by construction, so the marginal value of the diffusion stage should shrink under non-Gaussian impairments such as packet bursts or correlated interference; measuring that gap would isolate where the assumption binds.
  • The fixed 0.6 threshold in Coop-WD-eco could be replaced by a learned controller that trades AP for runtime continuously, since the weighting output already encodes distortion severity.
  • The same hierarchical principle could be applied to other shared representations, such as bird's-eye-view semantic maps or raw point-cloud features, where the distortion statistics differ and the weighting/denoising split would need rebalancing.
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

4 major / 4 minor

Summary. The paper proposes Coop-WD, a hierarchical feature-enhancement method for cooperative perception under V2V channel impairments. A self-supervised CAV-level weighting module is combined with a conditional diffusion-based pixel-level denoiser before feature fusion. The method is evaluated on the V2V4Real dataset with PointPillars and V2VNet under Rician fading, WINNER II, and a non-stationary V2V channel, including imperfect CSI, path-loss variation, and time-varying distortions. An efficient variant, Coop-WD-eco, selectively bypasses the denoiser under severe distortion to reduce runtime. The paper claims that Coop-WD consistently outperforms both weighting-only and denoising-only baselines across all SNR levels and channel types, and that Coop-WD-eco cuts computational cost by up to 50% with comparable accuracy.

Significance. If the results hold, the proposed integration of vehicle-level weighting and pixel-level diffusion denoising is a plausible and useful contribution for robust cooperative perception, and the breadth of channel conditions tested (including generalization to unseen channel models) is a clear strength. The use of a real-world cooperative-perception dataset and the inclusion of imperfect CSI, path loss, and time-varying distortions make the evaluation more realistic than many prior works. However, the central quantitative claim is overstated relative to the paper's own Table I, and the absence of statistical uncertainty reporting leaves even the non-contradicted cells difficult to assess. The efficient variant is an interesting practical addition, but its reported runtime savings are not consistent across all rows of Table V.

major comments (4)
  1. [Section V-B, Table I] The claim, repeated in the abstract and in Section V-B, that "Coop-WD consistently outperforms both Coop-W and Coop-D across all SNR levels" is directly contradicted by the reported numbers. At IoU=0.7 and 0 dB Rician fading, Coop-W achieves 0.280 while Coop-WD achieves 0.264; at IoU=0.7 and 0 dB non-stationary V2V, Coop-W scores 0.273 versus Coop-WD's 0.260; and at IoU=0.7 and 20 dB Rician fading, Coop-D scores 0.292 versus Coop-WD's 0.286. The paper needs to either modify the claim to a per-cell, non-universal statement or provide a corrected table, and the abstract and conclusion must be revised accordingly.
  2. [Section V (all results)] No error bars, number of seeds, or significance tests are reported for any of the AP values in Tables I-V or Figure 5. Several differences that are used to support the superiority of Coop-WD are very small, for example 0.483 vs. 0.481 at IoU=0.5 and 0 dB Rician fading, and 0.487 vs. 0.483 at IoU=0.5 and 0 dB non-stationary V2V. Without multiple seeds or paired evaluation across scenes, these differences could be run-to-run noise. The authors should report means and standard deviations over at least three seeds, or provide a paired significance test, for the main tables.
  3. [Section IV-A, Eq. (7)] The conditional diffusion forward process assumes that the corrupted feature at each step is a Gaussian linear interpolation between the clean transmitted feature x0 and the received distorted feature y, with an isotropic variance delta_t I. However, the actual feature distortion after zero-forcing detection, Rician or non-stationary fading, imperfect CSI, and time-varying noise is unlikely to be exactly Gaussian or state-independent. Because the reverse process is trained to invert this assumed corruption model, a mismatch could explain why denoising helps in some operating points and hurts in others. The authors should provide evidence that the modeled corruption matches the empirical feature-level distortion, for example by comparing the learned denoiser against a simple non-generative baseline under the same channel conditions, or by analyzing the distribution of feature residuals.
  4. [Section V-B, Table I and Section IV-B] The synergy narrative in Section IV-B states that weighting handles severe distortion while denoising handles moderate distortion, yet Table I shows Coop-WD losing to Coop-W alone at several low-SNR operating points (e.g., IoU=0.7, 0 dB Rician and 0 dB non-stationary V2V). This suggests the joint approach does not always combine the strengths of the two modules in the regime where weighting is claimed to dominate. The interpretation of the ablation needs to account for these cases explicitly rather than asserting consistent synergy.
minor comments (4)
  1. [Section V-D, second bullet] The text says "Coop-D falsely identifies some areas" in a comparison between Coop-W and Coop-WD; this should read "Coop-W" to match the comparison being made.
  2. [Section V-E, Table V] For the WINNER II channel at 20 dB, Coop-WD-eco has runtime 68.9 ms versus 69.0 ms for Coop-WD, so the claim of "approximately 10% improvement in computational efficiency" at higher SNR is not supported by this row.
  3. [Section V-A4 and Section V-E] The threshold of 0.6 used to activate the denoising module in Coop-WD-eco is introduced without any sensitivity analysis or justification; a brief study or discussion of how this threshold affects the accuracy-runtime trade-off would improve reproducibility.
  4. [General] The paper does not state the number of evaluation frames or the number of training/evaluation runs used to produce the reported AP values; this information should be included for reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Coop-WD's claimed gains are empirical, externally benchmarked, and not equivalent to its inputs by construction.

full rationale

I find no circular step in the paper's derivation chain. The proposed Coop-WD is a joint architecture whose weighting module is explicitly credited to the authors' prior work [23] and whose conditional diffusion formulation follows the external speech-enhancement framework [32]; neither citation is used to forbid alternatives or to define the claimed outcome. The central claims are empirical: the model is trained on a non-stationary V2V channel at SNR 15-20 dB and tested on Rician fading, WINNER II, and the same channel at different SNR levels, with AP measured against external baselines. No fitted parameter is renamed as a prediction, and the eco variant's threshold is an engineering control rule rather than a target result. The internal contradiction between the abstract's 'consistently outperforms' claim and several Table I cells (e.g., Coop-W beats Coop-WD at IoU=0.7, 0 dB Rician, 0.280 vs 0.264) is a correctness and reproducibility concern, not a circularity concern. The derivation is therefore self-contained, and the self-citation to [23] is a normal component citation rather than a load-bearing circular step.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical entity, force, particle, or conserved quantity. Its new elements are a learned architectural combination and a gating threshold, both counted as free parameters or domain assumptions above.

free parameters (6)
  • loss weight beta_coop = 0.1
    Coefficient for the detection loss in the joint training loss Eq. (23); chosen by hand, not derived.
  • loss weight beta_diffusion = 1
    Coefficient for the diffusion noise-prediction loss in Eq. (23); chosen by hand.
  • training SNR range = 15 to 20 dB
    Selected simulation condition for training; determines which distortion levels the diffusion model learns and how well it generalizes to lower SNRs.
  • diffusion noise schedule beta_t = 1e-4 to 0.035, 50 steps
    Forward-process noise schedule; a standard choice that affects denoising behavior and training cost.
  • inference denoising schedule = [0.0001, 0.001, 0.01, 0.05, 0.2, 0.35]
    Reverse-process fast sampling schedule taken from [32]; affects fidelity and runtime of the denoising module.
  • Coop-WD-eco gating threshold = 0.6
    Weighting output threshold that disables the denoising module; directly governs the claimed up to 50% runtime reduction and the accuracy trade-off.
assumptions (4)
  • domain assumption V2V channel corruption on intermediate features can be modeled by the conditional diffusion Gaussian form in Eq. (7).
    The entire pixel-level denoising module learns to reverse a linear Gaussian interpolation between clean feature x0 and received feature y; real equalized feature errors may be non-Gaussian, correlated, or state-dependent.
  • domain assumption The non-stationary V2V model [42] plus WINNER II and Rician fading are representative test channels for generalization.
    Training uses only the non-stationary V2V channel at 15 to 20 dB; the paper claims generalization by testing on unseen channel models, but all channels are simulated rather than measured over the air.
  • domain assumption The self-supervised weighting module [23] provides a scalar trust score that faithfully ranks distortion and can gate denoising.
    Coop-WD-eco switches off denoising when the trust score is below 0.6; if the score misranks distortion on unseen channels, the efficiency-accuracy trade-off is not guaranteed.
  • standard math The conditional diffusion ELBO and reverse-process formulas in Eqs. (11) to (22) are correct as stated.
    The paper cites CDiffuSE [32] and DDPM [45] and does not re-derive the equations; a transcription error would affect training and the reported gains.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Coop-WD: Cooperative Perception with Weighting and Denoising for Robust V2V Communication." pith.science (2026). https://pith.science/paper/HOOHV2JJ

@misc{pith2026250503528,
  author       = {Pith},
  title        = {Pith review of: Coop-WD: Cooperative Perception with Weighting and Denoising for Robust V2V Communication},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HOOHV2JJ}},
  note         = {Machine review of arXiv:2505.03528}
}
read the original abstract

Cooperative perception, leveraging shared information from multiple vehicles via vehicle-to-vehicle (V2V) communication, plays a vital role in autonomous driving to alleviate the limitation of single-vehicle perception. Existing works have explored the effects of V2V communication impairments on perception precision, but they lack generalization to different levels of impairments. In this work, we propose a joint weighting and denoising framework, Coop-WD, to enhance cooperative perception subject to V2V channel impairments. In this framework, the self-supervised contrastive model and the conditional diffusion probabilistic model are adopted hierarchically for vehicle-level and pixel-level feature enhancement. An efficient variant model, Coop-WD-eco, is proposed to selectively deactivate denoising to reduce processing overhead. Rician fading, non-stationarity, and time-varying distortion are considered. Simulation results demonstrate that the proposed Coop-WD outperforms conventional benchmarks in all types of channels. Qualitative analysis with visual examples further proves the superiority of our proposed method. The proposed Coop-WD-eco achieves up to 50% reduction in computational cost under severe distortion while maintaining comparable accuracy as channel conditions improve.

Figures

Figures reproduced from arXiv: 2505.03528 by the authors.

Figure 1
Figure 1. Cooperative perception via V2V communication. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. System architecture of cooperative perception with V2V communication. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Pixel-level diffusion and denoising processes for feature [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: The proposed joint weighting and denoising algorithm [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Performance under different path loss factors and time-varying disturbances. Disturbances are simulated with a fixed [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 7
Figure 7. Figure 7: Visualization examples of Coop-W (left) and Coop [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 6
Figure 6. Figure 6: Visualization examples of Coop-D (left) and the Coop [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. SimBEV2X: A Large-Scale Dataset and Data Generation Tool for Multi-Task Vehicle-to-Everything Cooperative Perception

    cs.CV 2026-07 conditional novelty 6.0 of 10

    SimBEV2X delivers a large-scale synthetic V2X dataset and generator with multi-task annotations, plus an attention fusion model that improves fused/lidar performance.

Reference graph

Works this paper leans on

46 extracted references · 23 canonical work pages · cited by 1 Pith paper

  1. [23]

    Self-supervised adaptive weighting for cooperative perception in v2v communications,

    C. Liu, J. Chen, Y . Chen, R. Payton, M. Riley, and S.-H. Yang, “Self-supervised adaptive weighting for cooperative perception in v2v communications,” IEEE Transactions on Intelligent V ehicles , vol. 9, no. 2, pp. 3569–3580, 2024

    work page 2024

  2. [1]

    Pointpillars: Fast encoders for object detection from point clouds,

    A. H. Lang, S. V ora, H. Caesar, L. Zhou, J. Yang, and O. Beijbom, “Pointpillars: Fast encoders for object detection from point clouds,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 12 697–12 705

    2019

  3. [2]

    Pv-rcnn: Point-voxel feature set abstraction for 3d object detection,

    S. Shi, C. Guo, L. Jiang, Z. Wang, J. Shi, X. Wang, and H. Li, “Pv-rcnn: Point-voxel feature set abstraction for 3d object detection,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 10 529–10 538

    2020

  4. [3]

    F-cooper: Feature based cooperative perception for autonomous vehicle edge computing system using 3d point clouds,

    Q. Chen, X. Ma, S. Tang, J. Guo, Q. Yang, and S. Fu, “F-cooper: Feature based cooperative perception for autonomous vehicle edge computing system using 3d point clouds,” in Proceedings of the 4th ACM/IEEE Symposium on Edge Computing , 2019, pp. 88–100

    2019

  5. [4]

    Opv2v: An open benchmark dataset and fusion pipeline for perception with vehicle-to- vehicle communication,

    R. Xu, H. Xiang, X. Xia, X. Han, J. Li, and J. Ma, “Opv2v: An open benchmark dataset and fusion pipeline for perception with vehicle-to- vehicle communication,” in 2022 International Conference on Robotics and Automation (ICRA) . IEEE, 2022, pp. 2583–2589

    2022

  6. [5]

    V2vnet: Vehicle-to-vehicle communication for joint perception and pre- diction,

    T. Wang, S. Manivasagam, M. Liang, B. Yang, W. Zeng, and R. Urtasun, “V2vnet: Vehicle-to-vehicle communication for joint perception and pre- diction,” in Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part II 16 . Springer, 2020, pp. 605–621

    work page 2020

  7. [6]

    V2x-vit: Vehicle-to-everything cooperative perception with vision transformer,

    R. Xu, H. Xiang, Z. Tu, X. Xia, M.-H. Yang, and J. Ma, “V2x-vit: Vehicle-to-everything cooperative perception with vision transformer,” in European conference on computer vision . Springer, 2022, pp. 107– 124

    2022

  8. [7]

    Cobevt: Cooper- ative bird’s eye view semantic segmentation with sparse transformers,

    R. Xu, Z. Tu, H. Xiang, W. Shao, B. Zhou, and J. Ma, “Cobevt: Cooper- ative bird’s eye view semantic segmentation with sparse transformers,” in Conference on Robot Learning . PMLR, 2023, pp. 989–1000

    work page 2023

Show all 46 references
  1. [8]

    When2com: Multi-agent perception via communication graph grouping,

    Y .-C. Liu, J. Tian, N. Glaser, and Z. Kira, “When2com: Multi-agent perception via communication graph grouping,” in Proceedings of the IEEE/CVF Conference on computer vision and pattern recognition , 2020, pp. 4106–4115

    2020

  2. [9]

    Where2comm: Communication-efficient collaborative perception via spatial confidence maps,

    Y . Hu, S. Fang, Z. Lei, Y . Zhong, and S. Chen, “Where2comm: Communication-efficient collaborative perception via spatial confidence maps,” Advances in neural information processing systems , vol. 35, pp. 4874–4886, 2022. 11

    2022

  3. [10]

    Who2com: Collaborative perception via learnable handshake communication,

    Y . Liu, J. Tian, C.-Y . Ma, N. Glaser, C.-W. Kuo, and Z. Kira, “Who2com: Collaborative perception via learnable handshake communication,” in 2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2020, pp. 6876–6883

    2020

  4. [11]

    On the road to 6g: Visions, requirements, key technologies, and testbeds,

    C.-X. Wang, X. You, X. Gao, X. Zhu, Z. Li, C. Zhang, H. Wang, Y . Huang, Y . Chen, H. Haas, J. S. Thompson, E. G. Larsson, M. D. Renzo, W. Tong, P. Zhu, X. Shen, H. V . Poor, and L. Hanzo, “On the road to 6g: Visions, requirements, key technologies, and testbeds,” IEEE Communi...

    2023

  5. [12]

    A vision of 6g wireless systems: Applications, trends, technologies, and open research problems,

    W. Saad, M. Bennis, and M. Chen, “A vision of 6g wireless systems: Applications, trends, technologies, and open research problems,” IEEE Network, vol. 34, no. 3, pp. 134–142, 2020

    2020

  6. [13]

    An introduction to deep learning for the physical layer,

    T. O’Shea and J. Hoydis, “An introduction to deep learning for the physical layer,” IEEE Transactions on Cognitive Communications and Networking, vol. 3, no. 4, pp. 563–575, 2017

    2017

  7. [14]

    Deep learning based detection for communications systems with radar interference,

    C. Liu, Y . Chen, and S.-H. Yang, “Deep learning based detection for communications systems with radar interference,” IEEE Transactions on V ehicular Technology, vol. 71, no. 6, pp. 6245–6254, 2022

    2022

  8. [15]

    Deep learning enabled semantic communication systems,

    H. Xie, Z. Qin, G. Y . Li, and B.-H. Juang, “Deep learning enabled semantic communication systems,” IEEE Transactions on Signal Pro- cessing, vol. 69, pp. 2663–2675, 2021

    2021

  9. [16]

    Knowledge distillation- based semantic communications for multiple users,

    C. Liu, Y . Zhou, Y . Chen, and S.-H. Yang, “Knowledge distillation- based semantic communications for multiple users,” IEEE Transactions on Wireless Communications , vol. 23, no. 7, pp. 7000–7012, 2024

    2024

  10. [17]

    Cooperative perception for 3d object detection in driving scenarios using infrastruc- ture sensors,

    E. Arnold, M. Dianati, R. de Temple, and S. Fallah, “Cooperative perception for 3d object detection in driving scenarios using infrastruc- ture sensors,” IEEE Transactions on Intelligent Transportation Systems , vol. 23, no. 3, pp. 1852–1864, 2020

    2020

  11. [18]

    Object classification using cnn-based fusion of vision and lidar in autonomous vehicle environment,

    H. Gao, B. Cheng, J. Wang, K. Li, J. Zhao, and D. Li, “Object classification using cnn-based fusion of vision and lidar in autonomous vehicle environment,” IEEE Transactions on Industrial Informatics , vol. 14, no. 9, pp. 4224–4231, 2018

    2018

  12. [19]

    Object-level perception sharing among connected vehicles,

    M. Ambrosin, I. J. Alvarez, C. Buerkle, L. L. Yang, F. Oboril, M. R. Sas- try, and K. Sivanesan, “Object-level perception sharing among connected vehicles,” in 2019 IEEE Intelligent Transportation Systems Conference (ITSC). IEEE, 2019, pp. 1566–1573

    2019

  13. [20]

    Distributed dynamic map fusion via federated learning for intelligent networked vehicles,

    Z. Zhang, S. Wang, Y . Hong, L. Zhou, and Q. Hao, “Distributed dynamic map fusion via federated learning for intelligent networked vehicles,” in 2021 IEEE International conference on Robotics and Automation (ICRA). IEEE, 2021, pp. 953–959

    2021

  14. [21]

    Learning for vehicle-to-vehicle cooperative perception under lossy communication,

    J. Li, R. Xu, X. Liu, J. Ma, Z. Chi, J. Ma, and H. Yu, “Learning for vehicle-to-vehicle cooperative perception under lossy communication,” IEEE Transactions on Intelligent V ehicles, vol. 8, no. 4, pp. 2650–2660, 2023

    2023

  15. [22]

    Co- operative perception with learning-based v2v communications,

    C. Liu, Y . Chen, J. Chen, R. Payton, M. Riley, and S.-H. Yang, “Co- operative perception with learning-based v2v communications,” IEEE Wireless Communications Letters, vol. 12, no. 11, pp. 1831–1835, 2023

    2023

  16. [24]

    Cooper: Cooperative perception for connected autonomous vehicles based on 3d point clouds,

    Q. Chen, S. Tang, Q. Yang, and S. Fu, “Cooper: Cooperative perception for connected autonomous vehicles based on 3d point clouds,” in 2019 IEEE 39th International Conference on Distributed Computing Systems (ICDCS). IEEE, 2019, pp. 514–524

    2019

  17. [25]

    Latency-aware collaborative perception,

    Z. Lei, S. Ren, Y . Hu, W. Zhang, and S. Chen, “Latency-aware collaborative perception,” in European Conference on Computer Vision . Springer, 2022, pp. 316–332

    2022

  18. [26]

    Keypoints-based deep feature fusion for cooperative vehicle detection of autonomous driving,

    Y . Yuan, H. Cheng, and M. Sester, “Keypoints-based deep feature fusion for cooperative vehicle detection of autonomous driving,” IEEE Robotics and Automation Letters , vol. 7, no. 2, pp. 3054–3061, 2022

    2022

  19. [27]

    Among us: Adversarially robust collaborative perception by consensus,

    Y . Li, Q. Fang, J. Bai, S. Chen, F. Juefei-Xu, and C. Feng, “Among us: Adversarially robust collaborative perception by consensus,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 186–195

    2023

  20. [28]

    Intercarrier interference self-cancellation scheme for ofdm mobile communication systems,

    Y . Zhao and S.-G. Haggman, “Intercarrier interference self-cancellation scheme for ofdm mobile communication systems,” IEEE transactions on Communications , vol. 49, no. 7, pp. 1185–1191, 2001

    2001

  21. [29]

    A non- stationary mobile-to-mobile channel model allowing for velocity and trajectory variations of the mobile stations,

    W. Dahech, M. P ¨atzold, C. A. Guti ´errez, and N. Youssef, “A non- stationary mobile-to-mobile channel model allowing for velocity and trajectory variations of the mobile stations,” IEEE Transactions on Wireless Communications, vol. 16, no. 3, pp. 1987–2000, 2017

    1987

  22. [30]

    Image super-resolution via iterative refinement,

    C. Saharia, J. Ho, W. Chan, T. Salimans, D. J. Fleet, and M. Norouzi, “Image super-resolution via iterative refinement,” IEEE transactions on pattern analysis and machine intelligence , vol. 45, no. 4, pp. 4713–4726, 2022

    2022

  23. [31]

    Diffir: Efficient diffusion model for image restoration,

    B. Xia, Y . Zhang, S. Wang, Y . Wang, X. Wu, Y . Tian, W. Yang, and L. Van Gool, “Diffir: Efficient diffusion model for image restoration,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 13 095–13 105

    2023

  24. [32]

    Conditional diffusion probabilistic model for speech enhancement,

    Y .-J. Lu, Z.-Q. Wang, S. Watanabe, A. Richard, C. Yu, and Y . Tsao, “Conditional diffusion probabilistic model for speech enhancement,” in ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2022, pp. 7402–7406

    2022

  25. [33]

    Cddm: Channel denoising diffusion models for wireless semantic communica- tions,

    T. Wu, Z. Chen, D. He, L. Qian, Y . Xu, M. Tao, and W. Zhang, “Cddm: Channel denoising diffusion models for wireless semantic communica- tions,” IEEE Transactions on Wireless Communications , 2024

    2024

  26. [34]

    Conditional denoising diffusion probabilistic models for data reconstruction enhancement in wireless communications,

    M. Letafati, S. Ali, and M. Latva-Aho, “Conditional denoising diffusion probabilistic models for data reconstruction enhancement in wireless communications,” IEEE Transactions on Machine Learning in Commu- nications and Networking , vol. 3, pp. 133–146, 2025

    2025

  27. [35]

    Diffusion model-aided data reconstruction in cell-free massive mimo downlink: A computation-aware approach,

    ——, “Diffusion model-aided data reconstruction in cell-free massive mimo downlink: A computation-aware approach,” IEEE Wireless Com- munications Letters , vol. 13, no. 11, pp. 3162–3166, 2024

    2024

  28. [36]

    Generative semantic communication: Diffusion models beyond bit recovery,

    E. Grassucci, S. Barbarossa, and D. Comminiello, “Generative semantic communication: Diffusion models beyond bit recovery,” arXiv preprint arXiv:2306.04321, 2023

    2023 arXiv

  29. [37]

    Commin: Semantic image communications as an inverse problem with inn-guided diffusion models,

    J. Chen, D. You, D. G ¨und¨uz, and P. L. Dragotti, “Commin: Semantic image communications as an inverse problem with inn-guided diffusion models,” in ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) . IEEE, 2024, pp. 6675–6679

    2024

  30. [38]

    Ssd: Single shot multibox detector,

    W. Liu, D. Anguelov, D. Erhan, C. Szegedy, S. Reed, C.-Y . Fu, and A. C. Berg, “Ssd: Single shot multibox detector,” in Computer Vision– ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11–14, 2016, Proceedings, Part I 14 . Springer, 2016, pp. 21–37

    2016

  31. [39]

    Pointnet: Deep learning on point sets for 3D classification and segmentation,

    R. Q. Charles, H. Su, M. Kaichun, and L. J. Guibas, “Pointnet: Deep learning on point sets for 3D classification and segmentation,” in Proc. IEEE Conf. Comput. Vision Pattern Recognit. (CVPR) , 2017, pp. 77–85

    2017

  32. [40]

    Focal loss for dense object detection,

    T.-Y . Lin, P. Goyal, R. Girshick, K. He, and P. Doll ´ar, “Focal loss for dense object detection,” in 2017 IEEE International Conference on Computer Vision (ICCV) , 2017, pp. 2999–3007

    2017

  33. [41]

    Winner ii channel models,

    P. Kyosti, “Winner ii channel models,” IST, Tech. Rep. IST- 4-027756 WINNER II D1.1.2 V1.2 , 2007. [Online]. Available: https://cir.nii.ac.jp/crid/1571698600936523008

    2007

  34. [42]

    A practical non-stationary channel model for vehicle-to-vehicle mimo commu- nications,

    W. Li, Q. Zhu, C.-X. Wang, F. Bai, X. Chen, and D. Xu, “A practical non-stationary channel model for vehicle-to-vehicle mimo commu- nications,” in 2020 IEEE Wireless Communications and Networking Conference (WCNC). IEEE, 2020, pp. 1–6

    2020

  35. [43]

    A novel 3d non-stationary wireless mimo channel simulator and hardware emulator,

    Q. Zhu, H. Li, Y . Fu, C.-X. Wang, Y . Tan, X. Chen, and Q. Wu, “A novel 3d non-stationary wireless mimo channel simulator and hardware emulator,” IEEE Transactions on Communications , vol. 66, no. 9, pp. 3865–3878, 2018

    2018

  36. [44]

    A novel 3d non-stationary vehicle-to-vehicle channel model and its spatial- temporal correlation properties,

    Q. Zhu, Y . Yang, X. Chen, Y . Tan, Y . Fu, C.-X. Wang, and W. Li, “A novel 3d non-stationary vehicle-to-vehicle channel model and its spatial- temporal correlation properties,” IEEE access, vol. 6, pp. 43 633–43 643, 2018

    2018

  37. [45]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” Advances in neural information processing systems , vol. 33, pp. 6840– 6851, 2020

    2020

  38. [46]

    V2v4real: A real-world large-scale dataset for vehicle-to-vehicle cooperative perception,

    R. Xu, X. Xia, J. Li, H. Li, S. Zhang, Z. Tu, Z. Meng, H. Xiang, X. Dong, R. Song et al. , “V2v4real: A real-world large-scale dataset for vehicle-to-vehicle cooperative perception,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, p...

    2023

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.