pith. sign in

arxiv: 2606.26398 · v2 · pith:PXKB5ZIPnew · submitted 2026-06-24 · 💻 cs.CV

DinoLink: A Token-Centric Representation Compression Framework for Bandwidth-Constrained Collaborative V2X Perception

Tianle Zhu , Haohua Que , Handong Yao , Hongyi Xu , Zhipeng Bao This is my paper

Pith reviewed 2026-07-01 06:22 UTC · model grok-4.3

classification 💻 cs.CV
keywords V2X perceptiontoken compressionsemantic communicationcollaborative inferencebandwidth efficiencyresidual vector quantizationsaliency-aware pruningnuScenes dataset
0
0 comments X

The pith

DinoLink replaces raw pixel streams with compressed semantic token indices to cut V2X transmission bitrate by 139 times while holding 32.8 percent mAP.

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

The paper presents DinoLink as a method to enable high-precision collaborative perception between vehicles and the cloud when network bandwidth is severely limited. It shifts from sending full images to sending only selected and quantized semantic tokens. A reader would care because this change directly addresses the data bottleneck that currently prevents reliable remote perception in real V2X settings. The approach uses a selector to drop background tokens and a quantizer to turn remaining features into short code indices plus position data. Simulations show the resulting system runs much faster on narrow-band links such as LoRa.

Core claim

DinoLink employs a dual-sparsity architecture: a saliency-aware selector prunes redundant background tokens, while a Residual Vector Quantization module collapses features into compact codebook indices. By transmitting only lightweight indices and positional priors, DinoLink achieves a 139× bitrate reduction compared to uncompressed transmission while maintaining a competitive 32.8% mAP on the nuScenes dataset. Deployment simulations further demonstrate a 34.5× acceleration in narrow-band environments, such as LoRa.

What carries the argument

Dual-sparsity architecture that combines a saliency-aware token selector with Residual Vector Quantization to convert image features into short codebook indices and position priors.

If this is right

  • V2X networks can deliver high-fidelity collaborative perception under tight bandwidth limits by sending indices instead of pixels.
  • Narrow-band links such as LoRa can support real-time remote perception with a measured 34.5 times speed-up.
  • The same frontend works across different perception backbones because no task-specific retraining is required.
  • Only code indices and position priors need to be transmitted, which directly lowers the data volume by the stated factor of 139.

Where Pith is reading between the lines

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

  • The same token-selection and quantization steps could be tested on other multi-agent sensing tasks such as drone fleets or roadside units.
  • If the learned codebook proves stable across datasets, the method could reduce the need to retrain perception models for each new compression setting.
  • Real-world channel noise and packet loss may alter the reported acceleration numbers obtained in simulation.

Load-bearing premise

The saliency selector and quantizer keep enough task-relevant information that the downstream perception model still works at the reported accuracy without any retraining or extra tuning.

What would settle it

Run the nuScenes validation set through the full DinoLink pipeline, transmit only the produced indices and priors, decode at the receiver, and measure whether mean average precision stays at or above 32.8 percent.

Figures

Figures reproduced from arXiv: 2606.26398 by Handong Yao, Haohua Que, Hongyi Xu, Tianle Zhu, Zhipeng Bao.

Figure 1
Figure 1. Figure 1: Overview of DinoLink. The edge vehicle extracts dense DINOv2 tokens, selects Top- [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Saliency-Aware Top-K Token Selection. (Left) Dense saliency map s t j derived from DINOv2 self-attention. (Right) Sparse token set Xt K after filtering redundant background. Normalized 2D positions pj are retained for server-side spatial reconstruction. it preserves semantically salient regions that are typically informative for query-driven downstream transformers. C. Token Quantization with Residual Vect… view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative detection results under different token ratios. Top rows are full-frame baselines; bottom rows keep only selected DINO patches. Higher [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: End-to-end latency comparison across diverse communication [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 4
Figure 4. Figure 4: Efficiency-accuracy trade-off analysis. The chart illustrates the [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Real-world experimental setup. (Left) Roof-mounted camera for real-time visual capture at the vehicle edge. (Right) The experimental vehicle platform used to validate DinoLink in physical V2X scenarios [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Per-frame detection metrics from our live vehicle-to-PC deployment [PITH_FULL_IMAGE:figures/full_fig_p007_7.png] view at source ↗
read the original abstract

High-precision remote perception is often hindered by the severe bandwidth constraints of Vehicle-to-Everything (V2X) networks. We propose \textit{DinoLink}, a token-centric compression framework that replaces raw pixel streaming with discrete semantic communication for vehicle-cloud collaborative inference. DinoLink employs a dual-sparsity architecture: a saliency-aware selector prunes redundant background tokens, while a Residual Vector Quantization (RVQ) module collapses features into compact codebook indices. By transmitting only lightweight indices and positional priors, DinoLink achieves a $139\times$ bitrate reduction compared to uncompressed transmission while maintaining a competitive 32.8\% mAP on the nuScenes dataset. Deployment simulations further demonstrate a $34.5\times$ acceleration in narrow-band environments, such as LoRa. Our results substantiate DinoLink as a robust, bandwidth-efficient frontend for high-fidelity remote perception in constrained V2X scenarios. The code is publicly available at https://github.com/UGA-MOBILITY-LAB/dino_link.

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 / 1 minor

Summary. The paper proposes DinoLink, a token-centric compression framework for bandwidth-constrained collaborative V2X perception. It replaces raw pixel streaming with discrete semantic communication via a dual-sparsity architecture: a saliency-aware selector that prunes redundant background tokens and a Residual Vector Quantization (RVQ) module that collapses features into compact codebook indices. By transmitting only lightweight indices and positional priors, the method claims a 139× bitrate reduction versus uncompressed transmission while achieving 32.8% mAP on nuScenes, plus 34.5× acceleration in narrow-band (e.g., LoRa) simulations. Public code is provided.

Significance. If the central claims are substantiated with clear evaluation protocols, the work could provide a practical frontend for remote perception under severe V2X bandwidth limits. Public code release supports reproducibility.

major comments (1)
  1. [Evaluation / Experimental Results] The evaluation protocol (likely §4 or the experimental section) must explicitly clarify whether the reported 32.8% mAP uses a frozen downstream perception head identical to the uncompressed baseline or after task-specific fine-tuning/adaptation on the compressed tokens. This detail is load-bearing: fine-tuning would show that the system can be adapted to the quantized stream rather than demonstrating that the saliency-aware selector + RVQ pipeline preserves sufficient semantics without retraining, directly affecting whether the 139× reduction claim demonstrates semantic preservation.
minor comments (1)
  1. [Abstract] Abstract reports headline numbers (139× reduction, 32.8% mAP, 34.5× acceleration) without referencing baselines, error bars, dataset splits, or ablations; these details should be summarized even in the abstract for immediate context.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for highlighting the need for explicit clarification on the evaluation protocol. This is a substantive point that affects interpretation of our semantic-preservation claims, and we will address it directly in revision.

read point-by-point responses

  1. Referee: [Evaluation / Experimental Results] The evaluation protocol (likely §4 or the experimental section) must explicitly clarify whether the reported 32.8% mAP uses a frozen downstream perception head identical to the uncompressed baseline or after task-specific fine-tuning/adaptation on the compressed tokens. This detail is load-bearing: fine-tuning would show that the system can be adapted to the quantized stream rather than demonstrating that the saliency-aware selector + RVQ pipeline preserves sufficient semantics without retraining, directly affecting whether the 139× reduction claim demonstrates semantic preservation.

    Authors: The reported 32.8% mAP is obtained using a frozen downstream perception head that is identical to the one used for the uncompressed baseline, with no task-specific fine-tuning or adaptation performed on the compressed tokens. This protocol was chosen precisely to isolate the semantic-preservation properties of the saliency-aware selector and RVQ pipeline. We will add an explicit statement of this protocol (including confirmation that the head remains frozen) to the experimental section and to the caption of the relevant result table. revision: yes

Circularity Check

0 steps flagged

No circularity detected; empirical results presented without reduction to inputs.

full rationale

The abstract and description report measured outcomes (139× bitrate reduction, 32.8% mAP on nuScenes) from the proposed saliency-aware selector and RVQ components, with no equations, fitted parameters renamed as predictions, self-citations, or derivation steps visible. No load-bearing claim reduces by construction to its own inputs; the results are presented as experimental validation on an external dataset rather than tautological restatements of the method.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no explicit free parameters, axioms, or invented entities; the central claim implicitly rests on the unstated premise that the chosen saliency metric and codebook generalize across scenes.

pith-pipeline@v0.9.1-grok · 5723 in / 1107 out tokens · 20746 ms · 2026-07-01T06:22:04.674979+00:00 · methodology

Review history (2 revisions) →

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

44 extracted references · 11 canonical work pages · 8 internal anchors

  1. [1]

    DINOv2: Learning Robust Visual Features without Supervision

    M. Oquab, T. Darcet, T. Moutakanni, H. V o, M. Szafraniec, V . Khali- dov, P. Fernandez, D. Haziza, F. Massa, A. El-Noubyet al., “Dinov2: Learning robust visual features without supervision,”arXiv preprint arXiv:2304.07193, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  2. [2]

    End-to-end object detection with transformers,

    N. Carion, F. Massa, G. Synnaeve, N. Usunier, A. Kirillov, and S. Zagoruyko, “End-to-end object detection with transformers,” in European conference on computer vision. Springer, 2020, pp. 213– 229

    2020

  3. [3]

    Deformable DETR: Deformable Transformers for End-to-End Object Detection

    X. Zhu, W. Su, L. Lu, B. Li, X. Wang, and J. Dai, “Deformable detr: Deformable transformers for end-to-end object detection,”arXiv preprint arXiv:2010.04159, 2020

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  4. [4]

    Deep learning with edge computing: A review,

    J. Chen and X. Ran, “Deep learning with edge computing: A review,” Proceedings of the IEEE, vol. 107, no. 8, pp. 1655–1674, 2019

    2019

  5. [5]

    Neurosurgeon: Collaborative intelligence between the cloud and mobile edge,

    Y . Kang, J. Hauswald, C. Gao, A. Rovinski, T. Mudge, J. Mars, and L. Tang, “Neurosurgeon: Collaborative intelligence between the cloud and mobile edge,”ACM SIGARCH Computer Architecture News, vol. 45, no. 1, pp. 615–629, 2017

    2017

  6. [6]

    Vehicular networking: A survey and tutorial on re- quirements, architectures, challenges, standards and solutions,

    G. Karagiannis, O. Altintas, E. Ekici, G. Heijenk, B. Jarupan, K. Lin, and T. Weil, “Vehicular networking: A survey and tutorial on re- quirements, architectures, challenges, standards and solutions,”IEEE communications surveys & tutorials, vol. 13, no. 4, pp. 584–616, 2011

    2011

  7. [7]

    The jpeg still picture compression standard,

    G. K. Wallace, “The jpeg still picture compression standard,”Com- munications of the ACM, vol. 34, no. 4, pp. 30–44, 1991

    1991

  8. [8]

    Overview of the h. 264/avc video coding standard,

    T. Wiegand, G. J. Sullivan, G. Bjontegaard, and A. Luthra, “Overview of the h. 264/avc video coding standard,”IEEE Transactions on circuits and systems for video technology, vol. 13, no. 7, pp. 560– 576, 2003

    2003

  9. [9]

    Understanding how image quality affects deep neural networks,

    S. Dodge and L. Karam, “Understanding how image quality affects deep neural networks,” in2016 eighth international conference on quality of multimedia experience (QoMEX). IEEE, 2016, pp. 1–6

    2016

  10. [10]

    Benchmarking Neural Network Robustness to Common Corruptions and Perturbations

    D. Hendrycks and T. Dietterich, “Benchmarking neural network ro- bustness to common corruptions and perturbations,”arXiv preprint arXiv:1903.12261, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1903

  11. [11]

    Distributed deep neural networks over the cloud, the edge and end devices,

    S. Teerapittayanon, B. McDanel, and H.-T. Kung, “Distributed deep neural networks over the cloud, the edge and end devices,” in2017 IEEE 37th international conference on distributed computing systems (ICDCS). IEEE, 2017, pp. 328–339

    2017

  12. [12]

    Experimental assessment of communication delay’s impact on connected automated vehicle speed volatility and energy consumption,

    W. Li, J. Rios-Torres, B. Wang, and Z. H. Khattak, “Experimental assessment of communication delay’s impact on connected automated vehicle speed volatility and energy consumption,” Communications in Transportation Research, vol. 4, p. 100136, 2024. [Online]. Available: https://www.sciencedirect.com/science/article/pii/ S2772424724000192

    2024

  13. [13]

    Dynamicvit: Efficient vision transformers with dynamic token sparsification,

    Y . Rao, W. Zhao, B. Liu, J. Lu, J. Zhou, and C.-J. Hsieh, “Dynamicvit: Efficient vision transformers with dynamic token sparsification,”Ad- vances in neural information processing systems, vol. 34, pp. 13 937– 13 949, 2021

    2021

  14. [14]

    Tokenlearner: What can 8 learned tokens do for images and videos?

    M. S. Ryoo, A. Piergiovanni, A. Arnab, M. Dehghani, and A. An- gelova, “Tokenlearner: What can 8 learned tokens do for images and videos?”arXiv preprint arXiv:2106.11297, 2021

    work page arXiv 2021

  15. [15]

    Soundstream: An end-to-end neural audio codec,

    N. Zeghidour, A. Luebs, A. Omran, J. Skoglund, and M. Tagliasac- chi, “Soundstream: An end-to-end neural audio codec,”IEEE/ACM Transactions on Audio, Speech, and Language Processing, vol. 30, pp. 495–507, 2021

    2021

  16. [16]

    High Fidelity Neural Audio Compression

    A. D ´efossez, J. Copet, G. Synnaeve, and Y . Adi, “High fidelity neural audio compression,”arXiv preprint arXiv:2210.13438, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  17. [17]

    Neural discrete representation learning,

    A. Van Den Oord, O. Vinyalset al., “Neural discrete representation learning,”Advances in neural information processing systems, vol. 30, 2017

    2017

  18. [18]

    Generating diverse high-fidelity images with vq-vae-2,

    A. Razavi, A. Van den Oord, and O. Vinyals, “Generating diverse high-fidelity images with vq-vae-2,”Advances in neural information processing systems, vol. 32, 2019

    2019

  19. [19]

    nuscenes: A multimodal dataset for autonomous driving,

    H. Caesar, V . Bankiti, A. H. Lang, S. V ora, V . E. Liong, Q. Xu, A. Krishnan, Y . Pan, G. Baldan, and O. Beijbom, “nuscenes: A multimodal dataset for autonomous driving,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 11 621–11 631

    2020

  20. [20]

    Vehicle-to-everything (v2x) services supported by lte-based systems and 5g,

    S. Chen, J. Hu, Y . Shi, Y . Peng, J. Fang, R. Zhao, and L. Zhao, “Vehicle-to-everything (v2x) services supported by lte-based systems and 5g,”IEEE communications standards magazine, vol. 1, no. 2, pp. 70–76, 2017

    2017

  21. [21]

    Wireless access in vehicular environments,

    W. Xiang, J. Gozalvez, Z. Niu, O. Altintas, and E. Ekici, “Wireless access in vehicular environments,”EURASIP Journal on Wireless Communications and Networking, vol. 2009, no. 1, p. 576217, 2009

    2009

  22. [22]

    V2vnet: Vehicle-to-vehicle communication for joint per- ception and prediction,

    T.-H. Wang, S. Manivasagam, M. Liang, B. Yang, W. Zeng, and R. Urtasun, “V2vnet: Vehicle-to-vehicle communication for joint per- ception and prediction,” inEuropean conference on computer vision. Springer, 2020, pp. 605–621

    2020

  23. [23]

    Dair-v2x: A large-scale dataset for vehicle- infrastructure cooperative 3d object detection,

    H. Yu, Y . Luo, M. Shu, Y . Huo, Z. Yang, Y . Shi, Z. Guo, H. Li, X. Hu, J. Yuanet al., “Dair-v2x: A large-scale dataset for vehicle- infrastructure cooperative 3d object detection,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 21 361–21 370

    2022

  24. [24]

    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,” in2022 International Conference on Robotics and Automation (ICRA). IEEE, 2022, pp. 2583–2589

    2022

  25. [25]

    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,” inEuropean conference on computer vision. Springer, 2022, pp. 107– 124

    2022

  26. [26]

    A wire- less collaborated inference acceleration framework for plant disease recognition,

    H. Zhu, X. Huang, H. Gao, M. Jiang, H. Que, and L. Mu, “A wire- less collaborated inference acceleration framework for plant disease recognition,” inInternational Conference on Intelligent Computing. Springer, 2025, pp. 331–341

    2025

  27. [27]

    Wireless collaborative inference acceleration based on distillation for weed detection and instance segmentation,

    R. Li, Y . Mo, R. Zhao, H. Gao, H. Que, and L. Mu, “Wireless collaborative inference acceleration based on distillation for weed detection and instance segmentation,” in2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2025, pp. 1847–1854

    2025

  28. [28]

    Roadside cross-camera vehicle tracking combining visual and spatial-temporal information for a cloud control system,

    B. Gao, Z. Li, D. Zhang, Y . Liu, J. Chen, and Z. Lv, “Roadside cross-camera vehicle tracking combining visual and spatial-temporal information for a cloud control system,”Journal of Intelligent and Connected Vehicles, vol. 7, no. 2, pp. 129–137, 2024

    2024

  29. [29]

    An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gellyet al., “An image is worth 16x16 words: Transformers for image recognition at scale,”arXiv preprint arXiv:2010.11929, 2020

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  30. [30]

    Swin transformer: Hierarchical vision transformer using shifted windows,

    Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, and B. Guo, “Swin transformer: Hierarchical vision transformer using shifted windows,” inProceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 10 012–10 022

    2021

  31. [31]

    Perception strategies in low- altitude transportation: Single aircraft autonomous system vs. aircraft- ground-cloud integration system,

    Y . Wang, K. Wang, J. Gong, and X. Qu, “Perception strategies in low- altitude transportation: Single aircraft autonomous system vs. aircraft- ground-cloud integration system,”Communications in Transportation Research, vol. 5, p. 100208, 2025. [Online]. Available: https: //www.sciencedirect.com/science/article/pii/S2772424725000484

    2025

  32. [32]

    Emerging properties in self-supervised vision trans- formers,

    M. Caron, H. Touvron, I. Misra, H. J ´egou, J. Mairal, P. Bojanowski, and A. Joulin, “Emerging properties in self-supervised vision trans- formers,” inProceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 9650–9660

    2021

  33. [33]

    Masked autoencoders are scalable vision learners,

    K. He, X. Chen, S. Xie, Y . Li, P. Doll ´ar, and R. Girshick, “Masked autoencoders are scalable vision learners,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 16 000–16 009

    2022

  34. [34]

    iBOT: Image BERT Pre-Training with Online Tokenizer

    J. Zhou, C. Wei, H. Wang, W. Shen, C. Xie, A. Yuille, and T. Kong, “ibot: Image bert pre-training with online tokenizer,”arXiv preprint arXiv:2111.07832, 2021

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  35. [35]

    Transfuser: Imitation with transformer-based sensor fusion for au- tonomous driving,

    K. Chitta, A. Prakash, B. Jaeger, Z. Yu, K. Renz, and A. Geiger, “Transfuser: Imitation with transformer-based sensor fusion for au- tonomous driving,”IEEE transactions on pattern analysis and machine intelligence, vol. 45, no. 11, pp. 12 878–12 895, 2022

    2022

  36. [36]

    Bev- former: learning bird’s-eye-view representation from lidar-camera via spatiotemporal transformers,

    Z. Li, W. Wang, H. Li, E. Xie, C. Sima, T. Lu, Q. Yu, and J. Dai, “Bev- former: learning bird’s-eye-view representation from lidar-camera via spatiotemporal transformers,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 47, no. 3, pp. 2020–2036, 2024

    2020

  37. [37]

    Lossy Image Compression with Compressive Autoencoders

    L. Theis, W. Shi, A. Cunningham, and F. Husz ´ar, “Lossy im- age compression with compressive autoencoders,”arXiv preprint arXiv:1703.00395, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  38. [38]

    End-to-end optimized image compression,

    J. Ball ´e, V . Laparra, and E. P. Simoncelli, “End-to-end optimized image compression,”arXiv preprint arXiv:1611.01704, 2016

    work page arXiv 2016

  39. [39]

    Joint autoregressive and hier- archical priors for learned image compression,

    D. Minnen, J. Ball ´e, and G. D. Toderici, “Joint autoregressive and hier- archical priors for learned image compression,” inAdvances in Neural Information Processing Systems, S. Bengio, H. Wallach, H. Larochelle, K. Grauman, N. Cesa-Bianchi, and R. Garnett, Eds., vol. 31. Curran Associates, Inc., 2018

    2018

  40. [40]

    Real-time adaptive image compression,

    O. Rippel and L. Bourdev, “Real-time adaptive image compression,” inProceedings of the 34th International Conference on Machine Learning - Volume 70, ser. ICML’17. JMLR.org, 2017, p. 2922–2930

    2017

  41. [41]

    Full Resolution Image Compression with Recurrent Neural Networks

    G. Toderici, D. Vincent, N. Johnston, S. Hwang, D. Minnen, J. Shor, and M. Covell, “Full resolution image compression with recurrent neural networks,”arXiv preprint arXiv:1608.05148, 08 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  42. [42]

    Learned image com- pression with discretized gaussian mixture likelihoods and attention modules,

    Z. Cheng, H. Sun, M. Takeuchi, and J. Katto, “Learned image com- pression with discretized gaussian mixture likelihoods and attention modules,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2020

    2020

  43. [43]

    High-fidelity generative image compression,

    F. Mentzer, G. Toderici, M. Tschannen, and E. Agustsson, “High-fidelity generative image compression,”arXiv preprint arXiv:2006.09965, 2020

    work page arXiv 2006

  44. [44]

    Taming transformers for high- resolution image synthesis,

    P. Esser, R. Rombach, and B. Ommer, “Taming transformers for high- resolution image synthesis,” 2020

    2020