Pith. sign in

REVIEW 3 major objections 5 minor 2 cited by

ICME 2025 Generalizable HDR and SDR Video Quality Measurement Grand Challenge

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The top-performing model in the ICME 2025 HDR/SDR VQA challenge was a no-reference vision-language model that beat VMAF and every other baseline.

desk verdict A solid, transparent challenge report whose 'state-of-the-art' claim outruns its statistics: the winning SROCC margin is within the noise floor of an 11-content test set. read the letter →

arxiv 2506.22790 v2 pith:7WBND7B4 submitted 2025-06-28 eess.IV cs.CVcs.MM

Yixu Chen , Bowen Chen , Hai Wei , Alan C. Bovik , Baojun Li , Wei Sun , Linhan Cao , Kang Fu
show 11 more authors
Dandan Zhu Jun Jia Menghan Hu Xiongkuo Min Guangtao Zhai Dounia Hammou Fei Yin Rafal Mantiuk Amritha Premkumar Prajit T Rajendran Vignesh V Menon
This is my paper · ORCID
classification eess.IVcs.CVcs.MM
keywords videoqualityassessmentHDRSDRno-referencefull-referencemultimodallargelanguagemodelgrandchallengeJODscores
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

This paper reports the outcomes of the ICME 2025 Grand Challenge on joint HDR and SDR video quality measurement. The organizers built a test bed from 31 open-source 8K HDR sources, each encoded in both HDR10 and SDR at nine bitrate-resolution levels, with subjective scores collected by pairwise comparisons and scaled to JOD units. Against this benchmark, the top no-reference model—a vision-language model fine-tuned with LoRA and fed HDR frames through an HDR-to-SDR preprocessing step—achieved an SROCC of 0.945 and PLCC of 0.943, outperforming the full-reference VMAF baseline (0.905) and several other submitted models. The paper's central claim is that a single dynamic-range-agnostic MLLM pipeline can rank compressed HDR and SDR video quality more accurately than established metrics, provided the evaluation protocol treats SDR and HDR as separate content when recovering subjective scores. That last condition matters: the paper itself shows that if SDR and HDR are scored jointly as the same source, most models' correlations drop sharply.

What carries the argument

The load-bearing machinery is the evaluation protocol paired with the winning model's preprocessing. The protocol derives ground-truth scores by converting pairwise human comparisons into continuous JOD values, and it recovers SDR and HDR scores as separate content; this separate-source recovery is what makes the reported rankings possible. On the model side, the key object is a multimodal large language model (an InternVL 2.5 vision-language model fine-tuned with LoRA), combined with FFmpeg Mobius tone mapping of HDR frames to SDR, linearization at 1000 nits, and sliding-window spatial crops of two-thirds of the shortest side. The preprocessing harmonizes HDR and SDR inputs into one distribution; the crops multiply the 360-video training set; and averaging six window predictions produces the final score.

What would settle it

Re-run the official test evaluation but recover the ground-truth JOD scores with SDR and HDR clips of the same source rated together in the same pairwise sessions (same-source recovery), then recompute SROCC for all models. If SLCV no longer beats P1204.3 or VMAF under that protocol, the state-of-the-art result holds only for the separate-source scoring choice.

Watch

Extended reading notes

Core claim

The central discovery is a benchmark result: in a standardized comparison on unseen HDR10 and SDR test content, the no-reference SLCV model reached an SROCC of 0.945 and PLCC of 0.943 across the full test set, with per-format scores of 0.933/0.953 SROCC for SDR and HDR10 respectively. This beats every baseline the organizers ran, including the standardized no-reference model P.1204.3 (0.925), the encoding-feature metric EQM (0.925), and the widely used full-reference VMAF (0.905), and it also edges out the best full-reference submissions. Four of the seven submitted models outperformed VMAF. The authors attribute the win to a recipe: convert HDR to SDR with Mobius tone mapping under a 1000-nit reference display assumption, sample spatial windows at two-thirds of the short side, and fine-tune InternVL 2.5 with LoRA for four epochs, averaging per-window predictions. The paper also documents that the ranking is protocol-dependent: when SDR and HDR versions of the same source are recovered jointly as one content in the subjective analysis, the SROCC of EQM falls from 0.92 to 0.68, P1204.3 from 0.92 to 0.62, VMAF from 0.90 to 0.55, while SLCV's correlation barely changes (0.94 to 0.96).

Load-bearing premise

The rankings hinge on the choice to recover SDR and HDR subjective scores as separate content; if the two formats are treated as the same source in the pairwise analysis, most models' correlations drop sharply and the reported winner may change.

Editorial extensions

If this is right

  • A no-reference model can outperform the full-reference VMAF on compressed HDR/SDR content, suggesting that reference signals may be unnecessary for this rating task.
  • The same MLLM pipeline, with tone-mapped inputs, can serve both HDR and SDR, which is a practical recipe for streaming services that must monitor mixed-format libraries.
  • The released benchmark gives future VQA models a fixed protocol and dataset, so new claims of generalization across dynamic ranges can be tested against the same numbers.
  • Because the protocol choice changes model rankings, challenge reports should report both separate-source and joint-recovery scores; the separate-source numbers alone may overstate cross-format generalization.

Reading between the lines

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

  • The near-flat SLCV correlation under joint recovery (0.94 to 0.96) suggests the MLLM may have implicitly learned HDR-versus-SDR preferences, not just within-format quality; a targeted test would be to fine-tune the same pipeline on jointly recovered labels and see if it can still rank cross-format pairs.
  • Because SLCV takes about 42 seconds per video on an L40S while EQM takes about 17 seconds, the performance gap may come with a deployment cost; a practical follow-up would distill the MLLM into a lightweight regressor or reduce the number of spatial windows.
  • The sharp VMAF drop under joint recovery suggests VMAF is not merely less accurate but miscalibrated across dynamic ranges; adding a dynamic-range-aware calibration layer could narrow the gap without a new architecture.
  • The challenge used a single encoder (libx265) and a single HDR-to-SDR conversion operator; testing the winning recipe on other codecs and tone-mapping operators would show whether the generalization claim extends beyond the benchmark's distortions.
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. The paper reports the design, results, and analysis of the ICME 2025 Grand Challenge on Generalizable HDR and SDR Video Quality Measurement. It describes the HDRSDR-VQA test data, the evaluation protocol based on SROCC/PLCC/KROCC/RMSE after nonlinear mapping, the seven submitted models across FR and NR tracks, and the comparison against P.1204.3, EQM, VMAF, and PSNR-Y. The reported top performer is the NR model SLCV (SROCC 0.945, PLCC 0.943), with four of seven submissions beating VMAF. The paper also includes an analysis showing that performance changes substantially when SDR and HDR scores are recovered jointly rather than separately.

Significance. If the reported results hold, the challenge provides a useful public benchmark and dataset for HDR/SDR VQA, and the SLCV result is a notable demonstration that a fine-tuned multimodal LLM with HDR-to-SDR preprocessing can rank compressed HDR/SDR videos competitively with dedicated metrics. The submission pipeline (Docker containers, held-out test set, released code and data) is a concrete strength that supports reproducibility. However, the headline 'state-of-the-art' claim rests on small SROCC gaps without statistical validation and on a single scoring protocol, so the significance is conditional.

major comments (3)
  1. [Section III, Table I] The claim that SLCV is state-of-the-art rests on SROCC differences (0.945 vs. 0.925 for EQM and P1204.3) that are not accompanied by confidence intervals or significance tests. Since the 198 test sequences come from only 11 source contents, the effective number of independent observations is small; content-clustered bootstrap or permutation tests are needed before the ranking can be considered established.
  2. [Section V] The official evaluation uses the scores in which SDR and HDR are rated as separate content, and the paper reports that under joint recovery SLCV improves to 0.96 while EQM drops from 0.92 to 0.68, P1204.3 from 0.92 to 0.62, and VMAF from 0.90 to 0.55. Because the challenge's stated goal is generalization across HDR and SDR, the protocol choice is load-bearing: different plausible scoring protocols materially change the ranking of methods. Please justify the choice (or report both), and discuss how the 'generalizable' claim depends on it.
  3. [Section II and Section III] The test set is drawn from the same HDRSDR-VQA dataset and the same encoding pipeline as the training set, so the reported 'generalizable' performance is only demonstrated within this distribution. No cross-dataset or domain-shift evaluation is provided; without it, the title claim of generalizability is not fully supported.
minor comments (5)
  1. [Section IV-C] The heading 'cvvdpMITransformer' appears to be a typo for 'cvvdpMlTransformer' (and similarly 'cvvdpMISaliency' should be 'cvvdpMlSaliency').
  2. [Section II] The footnote states that the challenge is sponsored by Amazon Prime Video, but no sponsor role or conflict-of-interest statement is provided; please clarify the sponsor's involvement.
  3. [Section V] The phrase 'For simplicity' does not justify the choice of scoring protocol; consider replacing it with a principled rationale or a reference to the companion dataset paper.
  4. [Table I] The inference-time footnote says 'unless noted otherwise,' but the CPU/GPU distinction is only in the text; label the table to indicate which rows used CPU vs. GPU.
  5. [Abstract and Section III] The abstract claims 'state-of-the-art performance,' but the paper only compares with the challenge baselines and submissions; consider qualifying to 'among the methods evaluated here.'

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the challenge ranking is an empirical evaluation on held-out test data, not a derivation from its own inputs.

full rationale

The paper reports a grand-challenge evaluation: participants' models are trained on a released training subset and evaluated on unseen test content (11 contents, 198 sequences). The central claim that SLCV achieves the highest SROCC (0.945) is a measured result against held-out labels, not a quantity that is fitted or defined in terms of the competitors' scores. The winning teams are co-authors of the report, and the authors' own EQM is used as a baseline, but neither fact makes the ranking circular: EQM is an independently published metric, and the comparison is computed on data not used to train the submissions. Section V's analysis of SDR/HDR score-recovery protocols is a sensitivity study of an explicit organizer choice, and the paper discloses that SLCV remains high (0.96) under the alternative joint-recovery protocol, so the headline result is not an artifact of a hidden circular definition. PLCC/RMSE are computed after a four-parameter nonlinear mapping fit on the private test set, which is a calibration caveat applying uniformly to all models and does not affect the primary SROCC ranking; the absence of confidence intervals is a statistical robustness concern, not circularity. No load-bearing step reduces to its own input, and no self-citation is invoked to forbidding alternatives or to justify the core claim.

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

The central claim is an empirical benchmark result, so the main dependencies are the subjective ground truth, the evaluation protocol, and the test set representativeness. The hand-picked parameters listed above are part of the winning model's pipeline and affect the reported performance. No new physical or conceptual entities are introduced.

free parameters (4)
  • Four-parameter logistic mapping (beta1 to beta4) = not disclosed
    Fitted on the private test set to map model predictions to MOS for PLCC and RMSE computation (Section II). Does not affect the primary SROCC metric, but is a hand-fitted transform on the evaluation data.
  • Reference display peak luminance = 1000 nits
    Chosen in the SLCV preprocessing pipeline for linearizing HDR signals before tone mapping (Section IV-A). This hand-picked value affects the winning model's input distribution and therefore its performance.
  • Mobius tone mapping parameter = 0.3
    Chosen in the SLCV FFmpeg HDR-to-SDR conversion (Section IV-A). A different value would change the preprocessed inputs and likely the SROCC.
  • LoRA fine-tuning hyperparameters for SLCV = rank=16, alpha=48, lr=2e-4, batch=24, 4 epochs
    Chosen by the SLCV team to fine-tune InternVL 2.5 (Section IV-A). These hyperparameters influence the reported 0.945 SROCC but are not fitted by this paper.
assumptions (4)
  • domain assumption The pairwise-comparison subjective study produces valid interval-scaled JOD scores for both HDR and SDR videos.
    The entire benchmark treats the JOD scores as ground truth. This is a standard assumption in VQA but is not independently verified in this paper.
  • ad hoc to paper SDR and HDR versions of the same source are treated as separate content in score recovery.
    Section V states this was a deliberate choice and shows that joint recovery changes SROCC substantially for several models. The rankings in Table I depend on this protocol assumption.
  • domain assumption The 11 held-out contents are representative enough to support generalizable performance claims.
    The test set contains 198 clips from 11 source videos. The paper makes a 'state-of-the-art' claim based on this small sample without statistical analysis.
  • domain assumption The submitted Docker containers faithfully implement the described models.
    The challenge required Docker submissions, but the paper does not provide third-party verification of the container contents or the reported inference times.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ICME 2025 Generalizable HDR and SDR Video Quality Measurement Grand Challenge." pith.science (2026). https://pith.science/paper/7WBND7B4

@misc{pith2026250622790,
  author       = {Pith},
  title        = {Pith review of: ICME 2025 Generalizable HDR and SDR Video Quality Measurement Grand Challenge},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7WBND7B4}},
  note         = {Machine review of arXiv:2506.22790}
}
read the original abstract

This paper reports IEEE International Conference on Multimedia \& Expo (ICME) 2025 Grand Challenge on Generalizable HDR and SDR Video Quality Measurement. With the rapid development of video technology, especially High Dynamic Range (HDR) and Standard Dynamic Range (SDR) contents, the need for robust and generalizable Video Quality Assessment (VQA) methods has become increasingly demanded. Existing VQA models often struggle to deliver consistent performance across varying dynamic ranges, distortion types, and diverse content. This challenge was established to benchmark and promote VQA approaches capable of jointly handling HDR and SDR content. In the final evaluation phase, five teams submitted seven models along with technical reports to the Full Reference (FR) and No Reference (NR) tracks. Among them, four methods outperformed VMAF baseline, while the top-performing model achieved state-of-the-art performance, setting a new benchmark for generalizable video quality assessment.

Figures

Figures reproduced from arXiv: 2506.22790 by the authors.

Figure 1
Figure 1. Scatter plots with non-linear mapping between prediction and JOD for all models on the test set. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Proposed NR-VQA pipeline based on InternVL 2.5. Tone-mapped HDR/SDR inputs are spatially cropped and passed [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Architecture of the CompressedVQA-HDR model. It extracts Swin Transformer features from reference and distorted [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Architecture overview of the proposed ColorVideoVDP-ML. (a) provides details on the ColorVideoVDP metric that is [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Architecture overview of the NR-CompressedVQA-HDR [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Architecture overview of the PhoenixVideo. [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: EQM NR vs JoD where SDR and HDR are recovered differently reference (FR) metrics do not support comparisons across different pixel formats. For those FR metrics that do support it, or for no-reference (NR) metrics, the correlations drop sharply when SDR and HDR are inc…

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. Efficient Face Image Quality Assessment via Self-training and Knowledge Distillation

    cs.CV 2025-07 conditional novelty 5.0 of 10

    A self-trained teacher and knowledge distillation produce a 1.18M-parameter face image quality model that nearly matches the 86.88M-parameter teacher on ICCV 2025 VQualA FIQA challenge data.

  2. Engagement Prediction of Short Videos with Large Multimodal Models

    cs.CV 2025-08 conditional novelty 4.0 of 10

    Large multimodal models, especially one that also processes audio, predict short-video engagement better than the prior feature-based baseline on the SnapUGC test set.

Reference graph

Works this paper leans on

26 extracted references · 21 canonical work pages · cited by 2 Pith papers

  1. [1]

    Hdrsdr-vqa: A subjective video quality dataset for hdr and sdr comparative evaluation,

    Bowen Chen, Cheng-han Lee, Yixu Chen, Zaixi Shang, Hai Wei, and Alan C. Bovik, “Hdrsdr-vqa: A subjective video quality dataset for hdr and sdr comparative evaluation,” 2025

    work page 2025

  2. [2]

    Avt-vqdb-uhd-2-hdr: An open 8k hdr source dataset for video quality research,

    Dominik Keller, Thomas Goebel, Valentin Siebenkees, Julius Prenzel, and Alexander Raake, “Avt-vqdb-uhd-2-hdr: An open 8k hdr source dataset for video quality research,” in 2024 16th International Confer- ence on Quality of Multimedia Experience (QoMEX) , 2024, pp. 186– 192

    work page 2024

  3. [3]

    Methodology for the subjective assessment of the quality of television pictures,

    B Series, “Methodology for the subjective assessment of the quality of television pictures,” Recommendation ITU-R BT , vol. 500, no. 13, 2012

    work page 2012

  4. [4]

    Active sampling for pairwise comparisons via approximate message passing and information gain maximization,

    Aliaksei Mikhailiuk, Clifford Wilmot, Maria Perez-Ortiz, Dingcheng Yue, and Rafał K Mantiuk, “Active sampling for pairwise comparisons via approximate message passing and information gain maximization,” in 2020 25th International Conference on Pattern Recognition (ICPR) . IEEE, 2021, pp. 2559–2566

    work page 2020

  5. [5]

    A practical guide and software for analysing pairwise comparison experiments,

    Maria Perez-Ortiz and Rafal K Mantiuk, “A practical guide and software for analysing pairwise comparison experiments,” arXiv preprint arXiv:1712.03686, 2017

    arXiv 2017

  6. [6]

    Study of subjective and objective quality assessment of video,

    Kalpana Seshadrinathan, Rajiv Soundararajan, Alan Conrad Bovik, and Lawrence K. Cormack, “Study of subjective and objective quality assessment of video,” IEEE Transactions on Image Processing , vol. 19, no. 6, pp. 1427–1441, 2010

    work page 2010

  7. [7]

    Multimodal large language model for hdr & sdr video quality measurement,

    Baojun Li, “Multimodal large language model for hdr & sdr video quality measurement,” in 2025 IEEE International Conference on Multimedia and Expo (ICME) , 2025

    work page 2025

  8. [8]

    Compressedvqa-hdr: Generalized full-reference and no-reference quality assessment models for com- pressed high dynamic range videos,

    Wei Sun, Linhan Cao, Kang Fu, Dandan Zhu, Jun Jia, Menghan Hu, Xiongkuo Min, and Guangtao Zhai, “Compressedvqa-hdr: Generalized full-reference and no-reference quality assessment models for com- pressed high dynamic range videos,” in 2025 IEEE International Conference on Multimedia and Expo (ICME) , 2025

    work page 2025

Show all 26 references
  1. [9]

    Colorvideovdp-ml: Visual difference predictor with a neural regressor for image and video quality prediction,

    Dounia Hammou, Fei Yin, and Rafal Mantiuk, “Colorvideovdp-ml: Visual difference predictor with a neural regressor for image and video quality prediction,” in 2025 IEEE International Conference on Multimedia and Expo (ICME) , 2025

    2025

  2. [10]

    Learn- ing quality from complexity and structure: A feature-fused xgboost model for video quality assessment,

    Amritha Premkumar, Prajit T Rajendran, and Vignesh V Menon, “Learn- ing quality from complexity and structure: A feature-fused xgboost model for video quality assessment,” in 2025 IEEE International Conference on Multimedia and Expo (ICME) , 2025

    2025

  3. [11]

    Bitstream-based model standard for 4k/uhd: Itu-t p.1204.3 – model details, evaluation, analysis and open source implementation,

    Rakesh Rao Ramachandra Rao, Steve G ¨oring, Werner Robitza, Alexan- der Raake, Bernhard Feiten, Peter List, and Ulf W ¨ustenhagen, “Bitstream-based model standard for 4k/uhd: Itu-t p.1204.3 – model details, evaluation, analysis and open source implementation,” in 2020 Twelfth ...

    2020

  4. [12]

    Encoder-quantization-motion-based video quality metrics,

    Yixu Chen, Zaixi Shang, Hai Wei, Yongjun Wu, and Sriram Sethuraman, “Encoder-quantization-motion-based video quality metrics,” in 2024 Picture Coding Symposium (PCS) , 2024, pp. 1–5

    2024

  5. [13]

    Vmaf: The journey continues,

    Zhi Li, Christos Bampis, Julie Novak, Anne Aaron, Kyle Swanson, Moorthy Anush, and Jan De Cock, “Vmaf: The journey continues,” 2018

    2018

  6. [14]

    Expanding performance boundaries of open-source multimodal models with model, data, and test-time scaling,

    Zhe Chen, Weiyun Wang, Yue Cao, Yangzhou Liu, Zhangwei Gao, Erfei Cui, Jinguo Zhu, Shenglong Ye, Hao Tian, Zhaoyang Liu, Lixin Gu, Xuehui Wang, Qingyun Li, Yimin Ren, Zixuan Chen, Jiapeng Luo, Jiahao Wang, Tan Jiang, Bo Wang, Conghui He, Botian Shi, Xingcheng Zhang, Han Lv, Yi...

    2025

  7. [15]

    Lora: Low-rank adaptation of large language models,

    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,” 2021

    2021

  8. [16]

    Deep learning based full-reference and no-reference quality assessment models for compressed ugc videos,

    Wei Sun, Tao Wang, Xiongkuo Min, Fuwang Yi, and Guangtao Zhai, “Deep learning based full-reference and no-reference quality assessment models for compressed ugc videos,” in IEEE International Conference on Multimedia & Expo Workshops , 2021, pp. 1–6

    2021

  9. [17]

    Image quality assessment: Unifying structure and texture similarity,

    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 , vol. 44, no. 5, pp. 2567–2581, 2020

    2020

  10. [18]

    Swin transformer: Hierarchical vision transformer using shifted windows,

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo, “Swin transformer: Hierarchical vision transformer using shifted windows,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2021, pp. 10012–10022

    2021

  11. [19]

    Challenge on quality assessment of compressed ugc videos,

    Haiqiang Wang, Gary Li, Shan Liu, and C.-C. Jay Kuo, “Challenge on quality assessment of compressed ugc videos,” Website, 2021, http: //ugcvqa.com/

    2021

  12. [20]

    Generalizable hdr & sdr video quality measurement grand challenge,

    “Generalizable hdr & sdr video quality measurement grand challenge,” Website, 2025, https://sites.google.com/view/icme25-vqm-gc/home

    2025

  13. [21]

    Colorvideovdp: A visual difference predictor for image, video and display distortions,

    Rafal K. Mantiuk, Param Hanji, Maliha Ashraf, Yuta Asano, and Alexandre Chapiro, “Colorvideovdp: A visual difference predictor for image, video and display distortions,” ACM Trans. Graph. , vol. 43, no. 4, July 2024

    2024

  14. [22]

    Siglip 2: Multilingual vision- language encoders with improved semantic understanding, localization, and dense features,

    Michael Tschannen, Alexey Gritsenko, Xiao Wang, Muhammad Ferjad Naeem, Ibrahim Alabdulmohsin, Nikhil Parthasarathy, Talfan Evans, Lucas Beyer, Ye Xia, Basil Mustafa, Olivier H ´enaff, Jeremiah Harmsen, Andreas Steiner, and Xiaohua Zhai, “Siglip 2: Multilingual vision- language...

    2025

  15. [23]

    Blind quality assessment for in-the-wild images via hierarchical feature fusion and iterative mixed database training,

    Wei Sun, Xiongkuo Min, Danyang Tu, Siwei Ma, and Guangtao Zhai, “Blind quality assessment for in-the-wild images via hierarchical feature fusion and iterative mixed database training,” IEEE Journal of Selected Topics in Signal Processing , vol. 17, no. 6, pp. 1178–1192, 2023

    2023

  16. [24]

    Green video complexity analysis for efficient encoding in adaptive video streaming,

    Vignesh V Menon, Christian Feldmann, Klaus Schoeffmann, Mo- hammed Ghanbari, and Christian Timmerer, “Green video complexity analysis for efficient encoding in adaptive video streaming,” in Proceed- ings of the First International Workshop on Green Multimedia Systems . June 20...

    2023

  17. [25]

    Image quality assessment: from error visibility to structural similarity,

    Zhou Wang, A.C. Bovik, H.R. Sheikh, and E.P. Simoncelli, “Image quality assessment: from error visibility to structural similarity,” IEEE Transactions on Image Processing , vol. 13, no. 4, pp. 600–612, 2004

    2004

  18. [26]

    Video quality assessment for resolution cross-over in live sports,

    Jingwen Zhu, Yixu Chen, Hai Wei, Sriram Sethuraman, and Yongjun Wu, “Video quality assessment for resolution cross-over in live sports,” in 2025 IEEE International Conference on Multimedia and Expo (ICME) , 2025

    2025

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.