Pith. sign in

REVIEW 3 major objections 5 minor 2 cited by

DiffVC-OSD: One-Step Diffusion-based Perceptual Neural Video Compression Framework

T0 review · 3 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read DiffVC-OSD claims that a one-step diffusion denoiser, guided by temporal context, can replace the 50-step denoising loop of diffusion video codecs, delivering state-of-the-art perceptual quality with 86.92% bitrate savings and about 20x fas

desk verdict First one-step diffusion NVC with a strong OSD-vs-MSD ablation, but the update rule is formally undefined at n=0 and the headline bitrate saving is overstated. read the letter →

arxiv 2508.07682 v1 pith:PJ4EELC6 submitted 2025-08-11 eess.IV cs.CV

Wenzhuo Ma , Zhenzhong Chen This is my paper · ORCID
classification eess.IVcs.CV
keywords neuralvideocompressionperceptualqualityone-stepdiffusiontemporalcontextadapterlatentrate-distortion-perceptiontrade-offcodingmodels
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

DiffVC-OSD claims that the iterative noise-then-denoise loop used by multi-step diffusion video codecs is unnecessary: feeding the noise-free reconstructed latent $\bar{y}_t$ directly into a single DDPM denoising pass, conditioned on temporal context, can produce perceptually enhanced reconstructions. A Temporal Context Adapter injects multi-level features extracted from the previously decoded frame and the latent into the denoising U-Net, while a final end-to-end finetuning stage optimizes rate, distortion, and perceptual losses together. The paper reports state-of-the-art perceptual compression on HEVC, MCL-JCV, and UVG, with an 86.92% bitrate reduction and about 20$\times$ faster P-frame decoding relative to its multi-step counterpart DiffVC. If correct, this removes the main latency and training-cost barriers that kept diffusion-based perceptual video compression away from practical use.

What carries the argument

The load-bearing mechanism is the one-step DDPM update in latent space: $\epsilon_\theta = \mathrm{Unet}(\bar{y}_t, c_t, n)$ with $c_t = \mathrm{Concat}(\bar{y}_t, C_t^0)$, followed by $\hat{y}_t = \frac{1}{\sqrt{\alpha_n}}(\bar{y}_t - \frac{1-\alpha_n}{\sqrt{1-\bar{\alpha}_n}}\epsilon_\theta)$. It replaces the multi-step noise-and-denoise chain with a single forward pass while keeping the compressed latent's structure. The Temporal Context Adapter makes that single step informative: it encodes multi-level temporal context from the previously decoded frame through ResBlocks and zero-initialized convolutions, with the first layer initialized by copying the U-Net's first-layer weights for the

What would settle it

Run the one-step pipeline with an explicitly indexed scheduler (for example, take the smallest nonzero timestep $n=1$) and compare LPIPS and DISTS on HEVC Class C against the 50-step variant. If the perceptual advantage disappears or reverses, the reported gains come from finetuning and adapter conditioning rather than from the one-step diffusion update itself. The same experiment settles whether Eq. (1) at the paper's stated $n=0$ is actually computable or relies on an unstated indexing convention.

Watch

Extended reading notes

Core claim

DiffVC-OSD's central claim is that a one-step diffusion update in the pretrained Stable Diffusion latent space is enough to perform the perceptual enhancement that prior diffusion codecs spread over 50 denoising iterations. The reconstructed latent $\bar{y}_t$ is concatenated with a large-scale temporal context $C_t^0$ to form the condition $c_t$, and the denoising U-Net predicts noise $\epsilon_\theta = \mathrm{Unet}(\bar{y}_t, c_t, n)$; the enhanced latent is then $\hat{y}_t = \frac{1}{\sqrt{\alpha_n}}(\bar{y}_t - \frac{1-\alpha_n}{\sqrt{1-\bar{\alpha}_n}}\epsilon_\theta)$. Because the starting point is the noise-free codec latent rather than pure noise, structural information is preserved

Load-bearing premise

The method assumes that a single denoising step applied directly to the compressed latent improves perceptual quality, and that the timestep reported as $n=0$ is a well-defined point on the noise schedule; the paper does not specify the scheduler indexing, and with standard indexing the update coefficients at $n=0$ are undefined.

Editorial extensions

If this is right

  • P-frame decoding drops from about 8.14 seconds to about 0.40 seconds on the same hardware, making diffusion-based perceptual video decoding roughly 20$\times$ faster.
  • The method preserves the reconstructed latent as the starting point, so structural information is not destroyed by adding noise, which is the key reason one step can suffice.
  • The ablation attributes an average 86.92% BD-rate penalty to the multi-step variant relative to the one-step method, implying equal perceptual quality at a much lower bitrate.
  • Because only one denoising step is needed, the whole framework can be finetuned end-to-end with a rate-distortion-perception loss; multi-step diffusion codecs could not afford the gradient accumulation.
  • The two additions beyond the bare one-step model each matter: the Temporal Context Adapter gives about 10.6% average perceptual BD-rate improvement, and end-to-end finetuning adds about 4.1%.

Reading between the lines

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

  • Beyond the paper, the same one-step recipe could apply to any learned codec whose latent lies near a pretrained diffusion prior; for image compression the temporal adapter would be dropped, and the latent alone would carry the condition.
  • The paper adopts $n=0$ for the denoising timestep without specifying the noise-scheduler indexing; at the canonical DDPM indexing the update coefficients are singular, so a reader should pin down the effective small-$n$ regime before building on the method.
  • If the 20$\times$ decode speedup transfers to lower-power hardware, diffusion-based perceptual compression becomes plausible for interactive video applications such as conferencing and game streaming, where multi-step latent diffusion was previously too slow.
  • A direct test of the mechanism's source: sweep $n$ while keeping everything else fixed and measure LPIPS/FID. If perceptual quality is flat in $n$, the gains come from finetuning and adapter conditioning rather than from the one-step diffusion update itself.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. DiffVC-OSD proposes a one-step diffusion-based perceptual video compression framework built on DiffVC. The reconstructed latent from the conditional coding pipeline is directly fed into a single DDPM denoising step guided by a Temporal Context Adapter, followed by end-to-end finetuning. The paper claims state-of-the-art perceptual quality on HEVC, MCL-JCV, and UVG, with about 20× faster decoding and an 86.92% bitrate reduction relative to its multi-step diffusion counterpart.

Significance. If the technical mechanism were sound, the idea of replacing multi-step iterative diffusion with a single, temporally conditioned refinement step would be a practically important contribution to perceptual neural video compression, cutting decoding latency substantially while preserving or improving perceptual metrics. The paper provides a useful ablation structure, evaluating the one-step vs. multi-step design, the temporal adapter, and end-to-end finetuning, across three standard datasets and several perceptual metrics. The main claims, however, rest on a one-step diffusion formulation that is not mathematically well-defined as stated, and on a headline bitrate-reduction number that is misread from the ablation table.

major comments (3)
  1. [§2.1, Eq. (1)] The central one-step update is not well-defined as a DDPM reverse step. The formula x̂_t = (1/√α_n)(ȳ_t − (1−α_n)/√(1−ā_n) ε_θ) is the standard reverse update for a noisy sample x_n at diffusion timestep n. The paper feeds the noise-free reconstructed latent ȳ_t directly, and Figure 5 sets n=0 as default. Under the usual DDPM schedule, n=0 is the clean end of the forward process (ā_0=1), making the coefficient (1−α_0)/√(1−ā_0) singular; under a 0-indexed schedule the input to the reverse step should be a slightly noised sample, not a clean latent. The paper never specifies the noise scheduler indices or justifies the operation. Because Table 2 attributes the main perceptual gains to this one-step update, the core mechanism is not reproducible as written. Please either specify a valid schedule and re-derive the update, or reframe the step as a learned single-pass refinement and avoid call
  2. [Abstract and §3.3, Table 2] The headline '86.92% bitrate reduction' is a misstatement. Table 2 uses DiffVC-OSD as the anchor. The row for the multi-step variant (Method D, MSD+TCA) has an average BD-rate of 86.92%, meaning that the multi-step method requires about 86.92% more bitrate than DiffVC-OSD at the same perceptual quality. This is not the same as DiffVC-OSD saving 86.92% of the bitrate; the relative bitrate saving is only 1 − 1/(1+0.8692) ≈ 46.5%. Please report the BD-rate number correctly or convert it to the actual percentage saving.
  3. [§3.2, Table 1] The introduction cites I2VC ('[2]') as a multi-step diffusion-based NVC, but I2VC is not included in any comparison. Given the claim of state-of-the-art perceptual compression performance across all test sets, the absence of I2VC is a significant omission. At minimum, the claim should be qualified to the compared methods, or I2VC should be added to Table 1 and the rate-perception curves.
minor comments (5)
  1. [§3.3, text around Table 2] The statement 'incorporating TCA yields an average improvement of 10.61%' does not match Table 2. The average BD-rate for Method B is 14.72 and for Method C is 4.11; that is a relative improvement of about 72%, not 10.61%. Please recompute and state the comparison correctly.
  2. [§3.3, text around Table 2] The sentence 'the comparison between Method C and Method D confirms that, compared to MSD, OSD provides an average 82.81% improvement' uses an absolute difference of average BD-rates (86.92 − 4.11) rather than a relative improvement. Please clarify whether the intended statistic is a relative gain or an absolute BD-rate gap.
  3. [§2.2, Figure 2] Notation is inconsistent: the text refers to temporal context as C0_t and C1_t, while Figure 2 uses C̄_t^0 and c_t^0...c_t^3. Please unify the notation and define what 'large-scale temporal context' means.
  4. [§3.3, Figure 5] The ablation for LoRA rank and timestep n does not label the vertical axis or state which BD-rate/metric is being plotted. Please give units and clarify whether lower is better.
  5. [Table 1 caption] Typo: 'RedandBlueindicate' should be 'Red and Blue indicate'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central SOTA claim is benchmarked against external codecs, and the self-comparisons are measured, not definitionally forced.

full rationale

The paper's central claims are (i) that DiffVC-OSD achieves state-of-the-art perceptual compression performance and (ii) that its one-step diffusion design yields large gains over its multi-step counterpart. Claim (i) is supported by Table 1, which compares against traditional codecs (HM, VTM), distortion-oriented NVCs (DCVC variants, SEVC), GAN-based NVCs (DVC-P, PLVC), and multi-step diffusion NVCs (DiffVC), with VTM-17.0 as anchor. This is external benchmarking, not a self-referential derivation. Claim (ii) is an ablation comparison in Table 2, where the 86.92% BD-rate and 20× decoding speedup are measured between the authors' own multi-step variant (Method D) and the proposed one-step system. Although the anchor for the BD-rate is the proposed method itself, this is an experimentally measured difference, not a parameter fitted to the target or a quantity equivalent to the input by construction; it may be a reporting choice but it is not circular. The paper does cite the authors' prior DiffVC work as the base framework and reuses its first seven training stages, but that is normal incremental research and is not used as an unverified uniqueness theorem or to forbid alternatives. The main technical concern—Eq. (1) with timestep n=0 in the default configuration is not a well-defined DDPM reverse update because the noise-schedule coefficients are undefined at n=0—is a correctness/reproducibility issue, not a circularity issue: the formula is imported from external DDPM literature and does not reduce to its own input. No load-bearing step equates a fitted input with a prediction, and no central result is equivalent by definition to an input. Therefore no circularity is present.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The method rests on domain assumptions about Stable Diffusion's latent space, the validity of a one-step denoising update on a clean latent, the compatibility of temporal context features with the frozen Unet, and the standard BD-rate evaluation methodology. The LoRA rank and timestep are tuned on a single test set, which is a source of potential selection bias. No new physical or conceptual entities are introduced; the Temporal Context Adapter is a neural network module.

free parameters (6)
  • LoRA rank r = 96
    Selected from an ablation over r values on HEVC Class C (Figure 5); affects capacity of the fine-tuned diffusion Unet.
  • Diffusion timestep n = 0
    Selected from ablation on HEVC Class C (Figure 5); the noise schedule index fed into the Unet in the one-step update (Eq. 1).
  • Loss weights w1, w2, w3 (stage 8) = 10.0, 1.0, 1.0
    Hand-set in Section 3.1 for the perceptual/fidelity fine-tuning stage.
  • Loss weights w4, w5, w6 (stage 9) = 0.8, 0.08, 0.08
    Hand-set in Section 3.1 for the rate-distortion-perception joint optimization.
  • Periodic weights wt = {0.5, 1.2, 0.5, 0.9}
    Hand-set period weights over four frames following DCVC-DC [11], Section 3.1.
  • Rate-distortion tradeoff lambda = {16, 48, 128, 384}
    Four rate points; standard practice following DiffVC [3], Section 3.1. These are evaluation anchors, not fitted, but they are hand-chosen operating points.
assumptions (5)
  • domain assumption Stable Diffusion V2.1-base's latent space and frozen VAE decoder are suitable for representing and reconstructing video frames.
    The method encodes frames into SD latents and decodes refined latents back to pixels (Section 2); the whole pipeline assumes this latent space is a good perceptual reconstruction space.
  • domain assumption A single denoising step applied to a noise-free latent ȳ_t (timestep n=0) yields a meaningful update, i.e., the one-step formula in Eq. (1) is well-defined and the network's epsilon prediction is useful.
    Section 2.1 defines the core operation; the default n=0 makes the standard DDPM coefficient 0/0 unless the schedule indexing is nonstandard, and the paper does not specify the schedule.
  • domain assumption The temporal context features C^0_t from DiffVC's conditional coding paradigm are compatible with the SD denoising Unet when concatenated with the latent.
    Section 2.2 describes TCA which maps these context features into the Unet; the paper assumes the context features can guide the diffusion prior without retraining the frozen parts.
  • standard math BD-rate/BD-metric computed relative to VTM-17.0 is a valid and standard way to compare perceptual compression across very different codecs, including those with non-overlapping rate-distortion curves.
    Section 3.2 uses BD-rate and BD-metric with VTM as anchor; many cells are N/A due to non-overlap, and BD-metric is used instead.
  • domain assumption The nine-stage training strategy inherits the first seven stages from DiffVC and is sufficient to reach the reported performance.
    Section 2.3 states stages 1-7 follow DiffVC, and stage 8/9 add LoRA and joint finetuning; the claim assumes DiffVC's training is a valid foundation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DiffVC-OSD: One-Step Diffusion-based Perceptual Neural Video Compression Framework." pith.science (2026). https://pith.science/paper/PJ4EELC6

@misc{pith2026250807682,
  author       = {Pith},
  title        = {Pith review of: DiffVC-OSD: One-Step Diffusion-based Perceptual Neural Video Compression Framework},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PJ4EELC6}},
  note         = {Machine review of arXiv:2508.07682}
}
abstract

In this work, we first propose DiffVC-OSD, a One-Step Diffusion-based Perceptual Neural Video Compression framework. Unlike conventional multi-step diffusion-based methods, DiffVC-OSD feeds the reconstructed latent representation directly into a One-Step Diffusion Model, enhancing perceptual quality through a single diffusion step guided by both temporal context and the latent itself. To better leverage temporal dependencies, we design a Temporal Context Adapter that encodes conditional inputs into multi-level features, offering more fine-grained guidance for the Denoising Unet. Additionally, we employ an End-to-End Finetuning strategy to improve overall compression performance. Extensive experiments demonstrate that DiffVC-OSD achieves state-of-the-art perceptual compression performance, offers about 20$\times$ faster decoding and a 86.92\% bitrate reduction compared to the corresponding multi-step diffusion-based variant.

Figures

Figures reproduced from arXiv: 2508.07682 by the authors.

Figure 1
Figure 1. The framework of DiffVC-OSD. and refines it in a single diffusion step, using both the large￾scale temporal context C 0 t and y¯t itself as conditions to enhance perceptual quality. Finally, the enhanced latent yˆt is decoded by the Stable Diffusion decoder to produce the reconstructed frame xˆt with rich details and a realistic appearance. In the following sections, we present detailed descriptions of the proposed … view at source ↗
Figure 2
Figure 2. The structure of TCA. 2.2 Temporal Context Adapter To enable the Denoising Unet to better leverage the condition ct by incorporating temporal context from the previously decoded frame—rather than relying solely on the current frame—we propose the Temporal Context Adapter (TCA), as illustrated in [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. The rate-perception curves of the DiffVC-OSD and other video compression methods on the MCL-JCV dataset [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: The ablation of lora position. 3.2 Experimental Results and Analysis To evaluate different categories of video compression methods, we select several representative approaches from each category for comparison. Traditional codecs include HM-16.25 [7] and VTM-17.0 [8]. …

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

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

  1. Group-of-Latents: Perceptual Video Compression at Extreme Bitrates via Masked Latent Generative Modeling

    eess.IV 2026-07 conditional novelty 7.0 of 10

    A generative video codec that transmits selected latent anchors and a text prompt, then synthesizes the rest with a diffusion transformer, reaches <0.005 bpp with strong perceptual quality.

  2. Generative Video Compression with Adaptive Score Distillation

    eess.IV 2026-07 conditional novelty 6.0 of 10

    A from-scratch, pixel-space video diffusion codec with a ground-truth-aligned gate on DMD gradients achieves one-step decoding and ~62–71% bitrate savings at matched LPIPS/FID over GLVC.

Reference graph

Works this paper leans on

25 extracted references · 22 canonical work pages · cited by 2 Pith papers

  1. [2]

    I 2vc: A unified framework for intra- & inter-frame video compression

    Meiqin Liu, Chenming Xu, Yukai Gu, Chao Yao, and Yao Zhao. I 2vc: A unified framework for intra- & inter-frame video compression. arXiv, abs/2405.14336, 2024

    arXiv 2024

  2. [1]

    Rethinking lossy com- pression: The rate-distortion-perception tradeoff

    Yochai Blau and Tomer Michaeli. Rethinking lossy com- pression: The rate-distortion-perception tradeoff. In Pro- ceedings of the International Conference on Machine Learning, volume 97, pages 675–685, 2019

    work page 2019

  3. [3]

    Diffusion-based per- ceptual neural video compression with temporal diffusion information reuse

    Wenzhuo Ma and Zhenzhong Chen. Diffusion-based per- ceptual neural video compression with temporal diffusion information reuse. arXiv, abs/2501.13528, 2025

    arXiv 2025

  4. [4]

    High-resolution im- age synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. High-resolution im- age synthesis with latent diffusion models. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10674–10685, 2022

    work page 2022

  5. [5]

    Denoising diffusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. In Advances in Neural Information Processing Systems, 2020

    2020

  6. [6]

    Adding conditional control to text-to-image diffusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. In IEEE/CVF International Conference on Computer Vision, pages 3813–3824, 2023

    work page 2023

  7. [7]

    https://vcgit.hhi.fraunhofer.de/ jvet/HM/, 2022

    HM-16.25. https://vcgit.hhi.fraunhofer.de/ jvet/HM/, 2022. Accessed: 2022-11-02

    work page 2022

  8. [8]

    https://vcgit.hhi.fraunhofer.de/ jvet/VVCSoftware_VTM/, 2022

    VTM-17.0. https://vcgit.hhi.fraunhofer.de/ jvet/VVCSoftware_VTM/, 2022. Accessed: 2022-11- 02

    work page 2022

Show all 25 references
  1. [9]

    Temporal context mining for learned video compres- sion

    Xihua Sheng, Jiahao Li, Bin Li, Li Li, Dong Liu, and Yan Lu. Temporal context mining for learned video compres- sion. IEEE Transactions on Multimedia, 25:7311–7322, 2023

    2023

  2. [10]

    Hybrid spatial-temporal en- tropy modelling for neural video compression

    Jiahao Li, Bin Li, and Yan Lu. Hybrid spatial-temporal en- tropy modelling for neural video compression. In ACM In- ternational Conference on Multimedia, pages 1503–1511, 2022

    2022

  3. [11]

    Neural video compression with diverse contexts

    Jiahao Li, Bin Li, and Yan Lu. Neural video compression with diverse contexts. In IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, pages 22616–22626, 2023

    2023

  4. [12]

    Neural video compression with feature modulation

    Jiahao Li, Bin Li, and Yan Lu. Neural video compression with feature modulation. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 26099– 26108, 2024

    2024

  5. [13]

    Aug- mented deep contexts for spatially embedded video coding

    Yifan Bian, Chuanbo Tang, Li Li, and Dong Liu. Aug- mented deep contexts for spatially embedded video coding. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2094–2104, 2025

  6. [14]

    DVC-P: deep video compression with perceptual optimizations

    Saiping Zhang, Marta Mrak, Luis Herranz, Marc Górriz Blanch, Shuai Wan, and Fuzheng Yang. DVC-P: deep video compression with perceptual optimizations. In Inter- national Conference on Visual Communications and Image Processing, pages 1–5, 2021

    2021

  7. [15]

    Percep- tual learned video compression with recurrent conditional GAN

    Ren Yang, Radu Timofte, and Luc Van Gool. Percep- tual learned video compression with recurrent conditional GAN. In Proceedings of International Joint Conference on Artificial Intelligence, pages 1537–1544, 2022

    2022

  8. [16]

    Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen

    Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. Lora: Low-rank adaptation of large language models. In International Conference on Learning Representations , 2022

    2022

  9. [17]

    Tianfan Xue, Baian Chen, Jiajun Wu, Donglai Wei, and William T. Freeman. Video enhancement with task- oriented flow. International Journal of Computer Vision, 127(8):1106–1125, 2019

    2019

  10. [18]

    Sullivan, Jens-Rainer Ohm, Woojin Han, and Thomas Wiegand

    Gary J. Sullivan, Jens-Rainer Ohm, Woojin Han, and Thomas Wiegand. Overview of the high efficiency video coding (HEVC) standard. IEEE Transactions on Circuits and Systems for Video Technology , 22(12):1649–1668, 2012

    2012

  11. [19]

    Haiqiang Wang, Weihao Gan, Sudeng Hu, Joe Yuchieh Lin, Lina Jin, Longguang Song, Ping Wang, Ioannis Kat- savounidis, Anne Aaron, and C.-C. Jay Kuo. MCL-JCV: A jnd-based H.264/A VC video quality assessment dataset. In IEEE International Conference on Image Processing, pages 150...

    2016

  12. [20]

    UVG dataset: 50/120fps 4k sequences for video codec analysis and development

    Alexandre Mercat, Marko Viitanen, and Jarno Vanne. UVG dataset: 50/120fps 4k sequences for video codec analysis and development. In Proceedings of the ACM Multimedia Systems Conference, pages 297–302, 2020

    2020

  13. [21]

    Efros, Eli Shecht- man, and Oliver Wang

    Richard Zhang, Phillip Isola, Alexei A. Efros, Eli Shecht- man, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. In IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, pages 586–595, 2018

    2018

  14. [22]

    Simoncelli

    Keyan Ding, Kede Ma, Shiqi Wang, and Eero P. Simoncelli. Image quality assessment: Unifying structure and texture similarity. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(5):2567–2581, 2022

    2022

  15. [23]

    Sutherland, Michael Arbel, and Arthur Gretton

    Mikolaj Binkowski, Danica J. Sutherland, Michael Arbel, and Arthur Gretton. Demystifying MMD gans. In Interna- tional Conference on Learning Representations, 2018

    2018

  16. [24]

    Gans trained by a two time-scale update rule converge to a local nash equi- librium

    Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, and Sepp Hochreiter. Gans trained by a two time-scale update rule converge to a local nash equi- librium. In Advances in Neural Information Processing Systems, pages 6626–6637, 2017

    2017

  17. [25]

    Mul- tiscale structural similarity for image quality assessment

    Zhou Wang, Eero P Simoncelli, and Alan C Bovik. Mul- tiscale structural similarity for image quality assessment. In Asilomar Conference on Signals, Systems & Computers, volume 2, pages 1398–1402, 2003.5

    2003

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.