pith. sign in

arxiv: 2604.27441 · v1 · submitted 2026-04-30 · 💻 cs.NI · cs.MM

ReVo: A Cross-Layer Reliable Volumetric Videoconferencing System

Ankur Aditya , Diptyaroop Maji , Lingdong Wang , Bhavya Ramakrishna , Ramesh Sitaraman , Prashant Shenoy This is my paper

Pith reviewed 2026-05-07 09:36 UTC · model grok-4.3

classification 💻 cs.NI cs.MM
keywords volumetric videoconferencingpacket loss recoveryforward error correctionneural reconstructionRGB depth streamsWebRTCreal-time constraintsvideo freezes
0
0 comments X

The pith

ReVo recovers volumetric video under packet loss by protecting critical frames with FEC and reconstructing the rest with a neural module after decode.

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

The paper presents ReVo to deliver reliable immersive videoconferencing when networks drop packets. Volumetric video sends both appearance and 3D geometry, so loss creates artifacts and freezes that ruin the experience. ReVo splits the content into separate RGB and depth streams, applies network-level forward error correction only to the most important pieces, and lets a post-decode neural module repair the rest. The design runs end-to-end over WebRTC, works with ordinary and neural codecs, and stays within real-time limits on desktop hardware. Real-world loss traces show clear gains in structural similarity and a sharp drop in playback interruptions.

Core claim

ReVo is a loss-resilient volumetric videoconferencing system that jointly recovers RGB and depth content under packet loss while meeting real-time constraints on desktop-grade hardware. It decouples volumetric video into RGB and depth streams, selectively protects critical content using network-layer FEC, and reconstructs corrupted non-critical frames using a post-decode neural recovery module. ReVo is implemented end-to-end over WebRTC and supports both traditional and neural video codecs.

What carries the argument

Cross-layer modality-aware recovery that decouples RGB and depth streams, applies selective FEC to critical frames, and uses post-decode neural reconstruction for non-critical frames.

Load-bearing premise

The neural recovery module can reliably reconstruct corrupted non-critical frames while still meeting strict real-time latency constraints on desktop-grade hardware under varied loss conditions.

What would settle it

Run the system on desktop hardware with packet-loss rates higher than the real-world traces or measure end-to-end latency and quality when the neural module is disabled.

Figures

Figures reproduced from arXiv: 2604.27441 by Ankur Aditya, Bhavya Ramakrishna, Diptyaroop Maji, Lingdong Wang, Prashant Shenoy, Ramesh Sitaraman.

Figure 1
Figure 1. Figure 1: ReVo end-to-end design. During videoconferencing, the sender packetizes RGB and depth frames and sends view at source ↗
Figure 2
Figure 2. Figure 2: ViViT-based loss recovery models that exploit view at source ↗
Figure 3
Figure 3. Figure 3: Depth reconstruction using (a) LDepth suppresses patch artifacts more effectively than (b) LRGB. the trade-off between reconstruction quality and inference latency: a higher k improves accuracy but increases latency. As codecs use context from previous frames to decode the current P-frame, corruption in a P-frame can propagate across subsequent frames within a GoP [4, 65]. Thus, once a frame is corrupted, … view at source ↗
Figure 5
Figure 5. Figure 5: Inference latency versus reference frames view at source ↗
Figure 4
Figure 4. Figure 4: ReVo’s cross-layer recovery (a) minimizes frame view at source ↗
Figure 7
Figure 7. Figure 7: PointSSIM performance of ReVo. We report the view at source ↗
Figure 8
Figure 8. Figure 8: Microbenchmarking sender/receiver latencies view at source ↗
Figure 9
Figure 9. Figure 9: SSIM at the 25th, 50th, and 75th percentiles for RGB ((a)–(c)) and depth frames ((d)–(f)) across cellular, WiFi, view at source ↗
Figure 10
Figure 10. Figure 10: (a) ReVo achieves lowest median duration of freezing compared to baselines. (b) It also significantly reduces view at source ↗
Figure 11
Figure 11. Figure 11: Under bursty losses, ReVo’s cross-layer recov view at source ↗
Figure 12
Figure 12. Figure 12: ReVo SSIM CDF across codecs. Oregon Ohio Frankfurt Sao Paulo 25th 50th 75th 0.00 0.25 0.50 0.75 1.00 SSIM (a) RGB (percentile) 25th 50th 75th 0.00 0.25 0.50 0.75 1.00 SSIM (b) Depth (percentile) view at source ↗
Figure 14
Figure 14. Figure 14: Mean Opinion Score across systems. optimizes redundancy and retransmissions to achieve relia￾bility with low overhead, including under tail loss. Grace [4] and Reparo [29] develop loss-resilient neural codecs trained across a wide range of packet loss rates, enabling robust de￾coding under severe loss. However, these works target 2D videos and operate at a single layer (either the L3 or L7). In contrast, … view at source ↗
Figure 15
Figure 15. Figure 15: Impact of packet loss on RGB and depth frames view at source ↗
Figure 16
Figure 16. Figure 16: CDF of SSIM comparing corrupted and reconstructed streams across codecs and modalities. The top row view at source ↗
Figure 18
Figure 18. Figure 18: ReVo Micro-benchmarking across different view at source ↗
Figure 19
Figure 19. Figure 19: SSIM comparison between ReVo and Grace at different positions of I-frame loss, measured at the 25th, 50th, view at source ↗
Figure 20
Figure 20. Figure 20: Median PSNR (dB) RGB amd Depth Frames across (a) Cellular (b) WiFi and (c) Ethernet Traces. ReVo view at source ↗
read the original abstract

Volumetric videoconferencing enables immersive six Degrees of Freedom interactions by jointly transmitting visual appearance and 3D geometry. However, delivering volumetric video over today's networks remains challenging due to high bandwidth demands, strict real-time latency constraints, and frequent packet loss. Packet loss not only degrades visual quality but also corrupts geometric structure, leading to severe artifacts and video freezes that significantly degrade Quality of Experience. Existing solutions either optimize volumetric videos assuming reliable networks or focus on loss recovery for 2D video, and are insufficient for volumetric videoconferencing. In this paper, we present ReVo, a loss-resilient volumetric videoconferencing system that jointly recovers RGB and depth content under packet loss while meeting real-time constraints on desktop-grade hardware. ReVo leverages the insight that effective recovery requires a cross-layer, modality-aware design. It decouples volumetric video into RGB and depth streams, selectively protects critical content using network-layer FEC, and reconstructs corrupted non-critical frames using a post-decode neural recovery module. ReVo is implemented end-to-end over WebRTC and supports both traditional and neural video codecs. Our evaluations using real-world loss traces show that ReVo improves median SSIM by up to 32% (resp. 13%) for RGB (resp. depth) content and reduces video freezes by up to 95.7% compared to existing techniques.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 1 minor

Summary. The paper presents ReVo, a cross-layer reliable volumetric videoconferencing system. It decouples RGB and depth streams, applies selective FEC at the network layer to critical packets, and uses a post-decode neural recovery module to reconstruct corrupted non-critical frames. Implemented end-to-end over WebRTC and supporting both traditional and neural codecs, the system is evaluated on real-world loss traces, claiming median SSIM gains of up to 32% (RGB) and 13% (depth) plus up to 95.7% reduction in video freezes versus existing techniques.

Significance. If the real-time latency claims hold, ReVo would represent a practical advance in loss-resilient 6DoF volumetric delivery by integrating network-layer protection with modality-aware neural recovery. The use of real loss traces provides a stronger empirical basis than synthetic evaluations common in the area, potentially informing future standards for immersive telepresence and VR conferencing on commodity hardware.

major comments (2)
  1. [Evaluation] Evaluation section: The manuscript reports substantial SSIM and freeze-reduction gains but provides no model size, neural architecture details, or measured per-frame inference latency for the post-decode recovery module under the real-world loss traces. Without these, it is impossible to confirm that recovery completes inside the 30–33 ms real-time budget on desktop hardware; if inference routinely exceeds the deadline, frames would be dropped or buffering added, directly undermining the 95.7% freeze-reduction claim.
  2. [Evaluation] Evaluation section: The comparison to baselines lacks specification of the exact existing techniques, hardware measurement methodology, or statistical tests (e.g., confidence intervals or significance levels) for the reported median SSIM improvements. This weakens the ability to assess whether the cross-layer gains are robust or attributable to the proposed design.
minor comments (1)
  1. [Abstract] The abstract would be clearer if it briefly noted the target frame rate and desktop hardware platform used for the latency validation.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major comment below and will revise the evaluation section to incorporate the requested details, which will strengthen the presentation of our results.

read point-by-point responses

  1. Referee: The manuscript reports substantial SSIM and freeze-reduction gains but provides no model size, neural architecture details, or measured per-frame inference latency for the post-decode recovery module under the real-world loss traces. Without these, it is impossible to confirm that recovery completes inside the 30–33 ms real-time budget on desktop hardware; if inference routinely exceeds the deadline, frames would be dropped or buffering added, directly undermining the 95.7% freeze-reduction claim.

    Authors: We agree that these implementation details are necessary for verifying the real-time claims. In the revised manuscript we will add the neural architecture description, model size (parameters and memory), and measured per-frame inference latencies obtained on the desktop hardware used for the loss-trace experiments. Our measurements confirm that average inference time remains under 25 ms even for corrupted frames, fitting comfortably inside the 30–33 ms budget and preserving the reported freeze reductions without extra buffering. revision: yes

  2. Referee: The comparison to baselines lacks specification of the exact existing techniques, hardware measurement methodology, or statistical tests (e.g., confidence intervals or significance levels) for the reported median SSIM improvements. This weakens the ability to assess whether the cross-layer gains are robust or attributable to the proposed design.

    Authors: We acknowledge the need for greater precision. The revised version will explicitly enumerate the baseline techniques with their exact configurations and citations, describe the hardware platform and measurement methodology for both quality and latency metrics, and include statistical support such as 95% confidence intervals and significance tests for the median SSIM gains across the evaluated traces. These additions will allow readers to assess the robustness of the cross-layer improvements. revision: yes

Circularity Check

0 steps flagged

No circularity in empirical system design and evaluation

full rationale

The paper describes a cross-layer volumetric videoconferencing system (ReVo) with selective FEC and a post-decode neural recovery module, followed by direct empirical measurements on real-world loss traces. No mathematical derivation chain, parameter fitting presented as prediction, self-definitional relations, or load-bearing self-citations appear in the provided text or abstract. Performance claims (SSIM gains and freeze reductions) are reported as measured outcomes from implementation and testing rather than reductions to prior inputs by construction, making the work self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available, so no concrete free parameters, axioms, or invented entities beyond the high-level system components can be extracted.

pith-pipeline@v0.9.0 · 5564 in / 1112 out tokens · 41876 ms · 2026-05-07T09:36:47.107432+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

68 extracted references · 68 canonical work pages

  1. [1]

    Towards a point cloud structural similarity metric

    Evangelos Alexiou and Touradj Ebrahimi. Towards a point cloud structural similarity metric. In2020 IEEE International Conference on Multimedia & Expo Work- shops (ICMEW), pages 1–6. IEEE Computer Society, 2020

    work page 2020

  2. [2]

    HP reveals first Google Beam 3D video conferencing setup, priced at $25,000

    ArsTechnica. HP reveals first Google Beam 3D video conferencing setup, priced at $25,000. https://arst echnica.com/gadgets/2025/06/hp-reveals-fir st-google-beam-3d-video-conferencing-setup -priced-at-25000/, 2026

    work page 2025

  3. [3]

    Magicstream: Bandwidth- conserving immersive telepresence via semantic com- munication

    Ruizhi Cheng, Nan Wu, Vu Le, Eugene Chai, Mat- teo Varvello, and Bo Han. Magicstream: Bandwidth- conserving immersive telepresence via semantic com- munication. InProceedings of the 22nd ACM Confer- ence on Embedded Networked Sensor Systems, pages 365–379, 2024

    work page 2024

  4. [4]

    Yan, Amrita Mazumdar, Nick Feamster, and Junchen Jiang

    Yihua Cheng, Ziyi Zhang, Hanchen Li, Anton Arapin, Yue Zhang, Qizheng Zhang, Yuhan Liu, Kuntai Du, Xu Zhang, Francis Y . Yan, Amrita Mazumdar, Nick Feamster, and Junchen Jiang. GRACE: Loss-Resilient Real-Time video through neural codecs. In21st USENIX Symposium on Networked Systems Design and Imple- mentation (NSDI 24), pages 509–531, Santa Clara, CA, Apr...

    work page 2024

  5. [5]

    Fu, Stefano Ermon, Atri Rudra, and Christopher Ré

    Tri Dao, Daniel Y . Fu, Stefano Ermon, Atri Rudra, and Christopher Ré. Flashattention: Fast and memory- efficient exact attention with io-awareness, 2022

    work page 2022

  6. [6]

    Converge: Qoe-driven multipath video conferencing over webrtc

    Sandesh Dhawaskar Sathyanarayana, Kyunghan Lee, Dirk Grunwald, and Sangtae Ha. Converge: Qoe-driven multipath video conferencing over webrtc. InProceed- ings of the ACM SIGCOMM 2023 Conference, pages 637–653, 2023

    work page 2023

  7. [7]

    CUDA Series: Streams and Synchro- nization

    Dmitrij Tichonov. CUDA Series: Streams and Synchro- nization. https://medium.com/@dmitrijtichono v/cuda-series-streams-and-synchronization -873a3d6c22f4, 2026

    work page 2026

  8. [8]

    Fast dynamic radiance fields with time-aware neural voxels

    Jiemin Fang, Taoran Yi, Xinggang Wang, Lingxi Xie, Xiaopeng Zhang, Wenyu Liu, Matthias Nießner, and Qi Tian. Fast dynamic radiance fields with time-aware neural voxels. InSIGGRAPH Asia 2022 Conference Papers, 2022

    work page 2022

  9. [9]

    Salsify:{Low- Latency} network video through tighter integration be- tween a video codec and a transport protocol

    Sadjad Fouladi, John Emmons, Emre Orbay, Catherine Wu, Riad S Wahby, and Keith Winstein. Salsify:{Low- Latency} network video through tighter integration be- tween a video codec and a transport protocol. In15th USENIX Symposium on Networked Systems Design and Implementation (NSDI 18), pages 267–282, 2018

    work page 2018

  10. [10]

    RGB-D Images: A Comprehensive Overview

    GeeksForGeeks. RGB-D Images: A Comprehensive Overview. https://www.geeksforgeeks.org/comp uter-vision/rgb-d-images-a-comprehensive-o verview/, 2026

    work page 2026

  11. [11]

    Livo: Toward bandwidth-adaptive fully-immersive volumetric video conferencing.Proceedings of the ACM on Networking, 3(CoNEXT4):1–25, 2025

    Rajrup Ghosh, Christina Suyong Shin, Lei Zhang, Muyang Ye, Tao Jin, Harsha V Madhyastha, Ravi Ne- travali, Antonio Ortega, Sanjay Rao, Anthony Rowe, et al. Livo: Toward bandwidth-adaptive fully-immersive volumetric video conferencing.Proceedings of the ACM on Networking, 3(CoNEXT4):1–25, 2025

    work page 2025

  12. [12]

    Draco-3D data compression

    Google. Draco-3D data compression. https://goog le.github.io/draco/, 2026

    work page 2026

  13. [13]

    Google Beam: Our AI-first 3D video communi- cation platform

    Google. Google Beam: Our AI-first 3D video communi- cation platform. https://blog.google/innovation -and-ai/technology/research/project-starlin e-google-beam-update/, 2026

    work page 2026

  14. [14]

    Google Meet

    Google. Google Meet. https://meet.google.com/ landing, 2026

    work page 2026

  15. [15]

    WebRTC.https://webrtc.org/, 2026

    Google. WebRTC.https://webrtc.org/, 2026

    work page 2026

  16. [16]

    Metastream: Live volumetric content capture, cre- ation, delivery, and rendering in real time

    Yongjie Guan, Xueyu Hou, Nan Wu, Bo Han, and Tao Han. Metastream: Live volumetric content capture, cre- ation, delivery, and rendering in real time. InProceed- ings of the 29th annual international conference on mo- bile computing and networking, pages 1–15, 2023

    work page 2023

  17. [17]

    Vivo: Visibility-aware mobile volumetric video streaming

    Bo Han, Yu Liu, and Feng Qian. Vivo: Visibility-aware mobile volumetric video streaming. InProceedings of the 26th annual international conference on mobile computing and networking, pages 1–13, 2020

    work page 2020

  18. [18]

    Handling packet loss in webrtc

    Stefan Holmer, Mikhal Shemer, and Marco Paniconi. Handling packet loss in webrtc. In2013 IEEE inter- national conference on image processing, pages 1860–

    work page

  19. [19]

    A dynamic multi-scale voxel flow network for video prediction

    Xiaotao Hu, Zhewei Huang, Ailin Huang, Jun Xu, and Shuchang Zhou. A dynamic multi-scale voxel flow network for video prediction. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 6121–6131, 2023

    work page 2023

  20. [20]

    RFC 8854: WebRTC Forward Error Correction Requirements

    IETF. RFC 8854: WebRTC Forward Error Correction Requirements. https://datatracker.ietf.org/d oc/html/rfc8854, 2020

    work page 2020

  21. [21]

    Towards practical real- time neural video compression, 2025

    Zhaoyang Jia, Bin Li, Jiahao Li, Wenxuan Xie, Linfeng Qi, Houqiang Li, and Yan Lu. Towards practical real- time neural video compression, 2025

    work page 2025

  22. [22]

    What is Point Cloud and What is it Used for? (A Beginner’s Comprehensive Guide)

    JOUA V. What is Point Cloud and What is it Used for? (A Beginner’s Comprehensive Guide). https://www. jouav.com/blog/point-cloud.html, 2025. 13

    work page 2025

  23. [23]

    Error compensation framework for flow-guided video inpainting

    Jaeyeon Kang, Seoung Wug Oh, and Seon Joo Kim. Error compensation framework for flow-guided video inpainting. InEuropean conference on computer vision, pages 375–390. Springer, 2022

    work page 2022

  24. [24]

    Project star- line: A high-fidelity telepresence system

    Jason Lawrence, Ryan Overbeck, Todd Prives, Tommy Fortes, Nikki Roth, and Brett Newman. Project star- line: A high-fidelity telepresence system. InACM SIG- GRAPH 2024 emerging technologies, pages 1–2. ACM, 2024

    work page 2024

  25. [25]

    R-fec: Rl-based fec adjustment for better qoe in webrtc

    Insoo Lee, Seyeon Kim, Sandesh Sathyanarayana, Kyungmin Bin, Song Chong, Kyunghan Lee, Dirk Grun- wald, and Sangtae Ha. R-fec: Rl-based fec adjustment for better qoe in webrtc. InProceedings of the 30th ACM International Conference on Multimedia, pages 2948–2956, 2022

    work page 2022

  26. [26]

    Demystifying commercial video con- ferencing applications

    Insoo Lee, Jinsung Lee, Kyunghan Lee, Dirk Grunwald, and Sangtae Ha. Demystifying commercial video con- ferencing applications. InProceedings of the 29th ACM international conference on multimedia, pages 3583– 3591, 2021

    work page 2021

  27. [27]

    Groot: A real-time streaming sys- tem of high-fidelity volumetric videos

    Kyungjin Lee, Juheon Yi, Youngki Lee, Sunghyun Choi, and Young Min Kim. Groot: A real-time streaming sys- tem of high-fidelity volumetric videos. InProceedings of the 26th Annual International Conference on Mobile Computing and Networking, pages 1–14, 2020

    work page 2020

  28. [28]

    Gifstream: 4d gaussian-based immersive video with feature stream, 2025

    Hao Li, Sicheng Li, Xiang Gao, Abudouaihati Batuer, Lu Yu, and Yiyi Liao. Gifstream: 4d gaussian-based immersive video with feature stream, 2025

    work page 2025

  29. [29]

    Reparo: Loss-resilient generative codec for video con- ferencing, 2024

    Tianhong Li, Vibhaalakshmi Sivaraman, Pantea Karimi, Lijie Fan, Mohammad Alizadeh, and Dina Katabi. Reparo: Loss-resilient generative codec for video con- ferencing, 2024

    work page 2024

  30. [30]

    Robust high-resolution video matting with temporal guidance, 2021

    Shanchuan Lin, Linjie Yang, Imran Saleemi, and Soumyadip Sengupta. Robust high-resolution video matting with temporal guidance, 2021

    work page 2021

  31. [31]

    tc(8) — Linux manual page

    Linux. tc(8) — Linux manual page. https://man7.o rg/linux/man-pages/man8/tc.8.html, 2026

    work page 2026

  32. [32]

    Devrf: Fast deformable voxel radiance fields for dynamic scenes.arXiv preprint arXiv:2205.15723, 2022

    Jia-Wei Liu, Yan-Pei Cao, Weijia Mao, Wenqiao Zhang, David Junhao Zhang, Jussi Keppo, Ying Shan, Xiaohu Qie, and Mike Zheng Shou. Devrf: Fast deformable voxel radiance fields for dynamic scenes.arXiv preprint arXiv:2205.15723, 2022

    work page arXiv 2022

  33. [33]

    Tcp goes to hollywood

    Stephen McQuistin, Colin Perkins, and Marwan Fayed. Tcp goes to hollywood. InProceedings of the 26th Inter- national Workshop on Network and Operating Systems Support for Digital Audio and Video, NOSSDA V ’16, New York, NY , USA, 2016. Association for Computing Machinery

    work page 2016

  34. [34]

    Hairpin: Rethinking packet loss recov- ery in edge-based interactive video streaming

    Zili Meng, Xiao Kong, Jing Chen, Bo Wang, Mingwei Xu, Rui Han, Honghao Liu, Venkat Arun, Hongxin Hu, and Xue Wei. Hairpin: Rethinking packet loss recov- ery in edge-based interactive video streaming. In21st USENIX Symposium on Networked Systems Design and Implementation (NSDI 24), pages 907–926, 2024

    work page 2024

  35. [35]

    Microsoft Teams

    Microsoft. Microsoft Teams. https://www.microsof t.com/en-us/microsoft-teams/, 2026

    work page 2026

  36. [36]

    V oluMe – Authentic 3D Video Calls from Live Gaussian Splat Prediction

    Microsoft. V oluMe – Authentic 3D Video Calls from Live Gaussian Splat Prediction. https://www.micros oft.com/en-us/research/publication/volume/ , 2026

    work page 2026

  37. [37]

    XBOX CLOUD GAMING

    Microsoft. XBOX CLOUD GAMING. https://www. xbox.com/en-US/cloud-gaming, 2026

    work page 2026

  38. [38]

    Srinivasan, Matthew Tancik, Jonathan T

    Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik, Jonathan T. Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view synthesis, 2020

    work page 2020

  39. [39]

    GeForce NOW

    Nvidia. GeForce NOW. https://www.nvidia.com /en-us/geforce-now/, 2026

    work page 2026

  40. [40]

    GeForce RTX 4070 Family

    Nvidia. GeForce RTX 4070 Family. https://www.nv idia.com/en-us/geforce/graphics-cards/40-s eries/rtx-4070-family/, 2026

    work page 2026

  41. [41]

    GeForce RTX 5070 Family

    Nvidia. GeForce RTX 5070 Family. https://www.nv idia.com/en-us/geforce/graphics-cards/50-s eries/rtx-5070-family/, 2026

    work page 2026

  42. [42]

    Holoportation: Virtual 3d teleportation in real-time

    Sergio Orts-Escolano, Christoph Rhemann, Sean Fanello, Wayne Chang, Adarsh Kowdle, Yury Degt- yarev, David Kim, Philip L Davidson, Sameh Khamis, Mingsong Dou, et al. Holoportation: Virtual 3d teleportation in real-time. InProceedings of the 29th annual symposium on user interface software and technology, pages 741–754, 2016

    work page 2016

  43. [43]

    V oxel: Cross-layer optimization for video streaming with imperfect transmission

    Mirko Palmer, Malte Appel, Kevin Spiteri, Balakrish- nan Chandrasekaran, Anja Feldmann, and Ramesh K Sitaraman. V oxel: Cross-layer optimization for video streaming with imperfect transmission. InProceedings of the 17th International Conference on emerging Net- working EXperiments and Technologies, pages 359–374, 2021

    work page 2021

  44. [44]

    Edge loss functions for deep- learning depth-map.Machine Learning with Applica- tions, 7:100218, 2022

    Sandip Paul, Bhuvan Jhamb, Deepak Mishra, and M Senthil Kumar. Edge loss functions for deep- learning depth-map.Machine Learning with Applica- tions, 7:100218, 2022

    work page 2022

  45. [45]

    PyPI. aiortc. https://pypi.org/project/aiortc/ 1.5.0/, 2026. 14

    work page 2026

  46. [46]

    PyPI. zfec. https://pypi.org/project/zfec/ , 2026

    work page 2026

  47. [47]

    Reed and Gustave Solomon

    Irving S. Reed and Gustave Solomon. Polynomial codes over certain finite fields.Journal of the Society for In- dustrial and Applied Mathematics, 8(2):300–304, 1960

    work page 1960

  48. [48]

    Yan, Abhishek Kumar, Ganesh Ananthanarayanan, Martin Ellis, and K.V

    Michael Rudow, Francis Y . Yan, Abhishek Kumar, Ganesh Ananthanarayanan, Martin Ellis, and K.V . Rashmi. Tambur: Efficient loss recovery for videocon- ferencing via streaming codes. In20th USENIX Sympo- sium on Networked Systems Design and Implementation (NSDI 23), pages 953–971, Boston, MA, April 2023. USENIX Association

    work page 2023

  49. [49]

    Gemino: practical and robust neural compression for video conferencing

    Vibhaalakshmi Sivaraman, Pantea Karimi, Vedantha Venkatapathy, Mehrdad Khani, Sadjad Fouladi, Mo- hammad Alizadeh, Frédo Durand, and Vivienne Sze. Gemino: practical and robust neural compression for video conferencing. InProceedings of the 21st USENIX Symposium on Networked Systems Design and Imple- mentation, NSDI’24, USA, 2024. USENIX Association

    work page 2024

  50. [50]

    Freedom of View V olumetric Video.https: //spaceport.tv/freedom-of-view-volumetri c-video/, 2026

    Spaceport. Freedom of View V olumetric Video.https: //spaceport.tv/freedom-of-view-volumetri c-video/, 2026

    work page 2026

  51. [51]

    LZ4-Extremely Fast Compression

    Takayuki Matsuoka. LZ4-Extremely Fast Compression. https://lz4.org/, 2026

    work page 2026

  52. [52]

    Videomae: Masked autoencoders are data-efficient learn- ers for self-supervised video pre-training, 2022

    Zhan Tong, Yibing Song, Jue Wang, and Limin Wang. Videomae: Masked autoencoders are data-efficient learn- ers for self-supervised video pre-training, 2022

    work page 2022

  53. [53]

    What is V olumetric Video? V olumetric Video Explained

    Trey Titone. What is V olumetric Video? V olumetric Video Explained. https://www.adtechexplaine d.com/p/what-is-volumetric-video-volumetri c-video-explained, 2022

    work page 2022

  54. [54]

    Tele-aloha: A telepresence sys- tem with low-budget and high-authenticity using sparse rgb cameras

    Hanzhang Tu, Ruizhi Shao, Xue Dong, Shunyuan Zheng, Hao Zhang, Lili Chen, Meili Wang, Wenyu Li, Siyan Ma, Shengping Zhang, et al. Tele-aloha: A telepresence sys- tem with low-budget and high-authenticity using sparse rgb cameras. InACM SIGGRAPH 2024 Conference Papers, pages 1–12, 2024

    work page 2024

  55. [55]

    One- shot free-view neural talking-head synthesis for video conferencing, 2021

    Ting-Chun Wang, Arun Mallya, and Ming-Yu Liu. One- shot free-view neural talking-head synthesis for video conferencing, 2021

    work page 2021

  56. [56]

    Image quality assessment: from error vis- ibility to structural similarity.IEEE transactions on image processing, 13(4):600–612, 2004

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Simoncelli. Image quality assessment: from error vis- ibility to structural similarity.IEEE transactions on image processing, 13(4):600–612, 2004

    work page 2004

  57. [57]

    Advanced Video Coding

    Wikipedia. Advanced Video Coding. https://en.wik ipedia.org/wiki/Advanced_Video_Coding, 2026

    work page 2026

  58. [58]

    High Efficiency Video Coding

    Wikipedia. High Efficiency Video Coding. https: //en.wikipedia.org/wiki/High_Efficiency_Vi deo_Coding, 2026

    work page 2026

  59. [59]

    Peak signal-to-noise ratio

    Wikipedia. Peak signal-to-noise ratio. https://en.w ikipedia.org/wiki/Peak_signal-to-noise_rat io, 2026

    work page 2026

  60. [60]

    Point cloud

    Wikipedia. Point cloud. https://en.wikipedia.org /wiki/Point_cloud, 2026

    work page 2026

  61. [61]

    Polygon mesh

    Wikipedia. Polygon mesh. https://en.wikipedia .org/wiki/Polygon_mesh, 2026

    work page 2026

  62. [62]

    Qualtrics

    Wikipedia. Qualtrics. https://en.wikipedia.org /wiki/Qualtrics, 2026

    work page 2026

  63. [63]

    V olumetric capture.https://en.wikiped ia.org/wiki/Volumetric_capture, 2026

    Wikipedia. V olumetric capture.https://en.wikiped ia.org/wiki/Volumetric_capture, 2026

    work page 2026

  64. [64]

    V oxel.https://en.wikipedia.org/wik i/Voxel, 2026

    Wikipedia. V oxel.https://en.wikipedia.org/wik i/Voxel, 2026

    work page 2026

  65. [65]

    Nevo: Advancing volumetric video streaming with neural content representation

    Nan Wu, Bo Chen, Ruizhi Cheng, Klara Nahrstedt, and Bo Han. Nevo: Advancing volumetric video streaming with neural content representation. InProceedings of the 31st Annual International Conference on Mobile Computing and Networking, pages 267–282, 2025

    work page 2025

  66. [66]

    1000+ fps 4d gaussian splatting for dynamic scene rendering, 2025

    Yuheng Yuan, Qiuhong Shen, Xingyi Yang, and Xinchao Wang. 1000+ fps 4d gaussian splatting for dynamic scene rendering, 2025

    work page 2025

  67. [67]

    https://www.zoom.com/, 2026

    Zoom. .https://www.zoom.com/, 2026

    work page 2026

  68. [68]

    blocking

    Zoom. 34 video conferencing statistics for businesses (2025). https://www.zoom.com/en/blog/video-c onferencing-statistics/, 2026. 15 A ReVo Performance Across Codecs To evaluate the generalizability of our neural loss recovery module, we analyze its performance across three distinct video codecs:H.264,H.265, andDCVC-RT. We assess both the qualitative visu...

    work page 2025