Pith. sign in

REVIEW 3 major objections 5 minor 60 references

3DGSI-Assessor: A Large-Scale Dataset and An LMM-based Method for 3D Gaussian Splatting Image Quality Assessment

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 A single LMM-based metric scores 3DGS image quality in three dimensions and outperforms more than 30 existing metrics on a new 15,200-image benchmark.

desk verdict A genuinely useful multi-dimensional 3DGS IQA dataset with a plausible LMM metric, but the in-domain SOTA claim is weakened by a train/test split that shares 3DGS model identities; worth reviewing with a required model-disjoint re-evaluation. read the letter →

arxiv 2608.03279 v1 pith:34DOGJA5 submitted 2026-08-04 cs.CV

Yuke Xing , Jiarui Wang , William Gordon , Zhu Li , Guangtao Zhai , Yiling Xu This is my paper
classification cs.CV
keywords 3DGaussiansplattingimagequalityassessmentlargemultimodalmodelmulti-dimensionalMOS3DGScompressionnovelviewsynthesisbenchmarkdatasetLoRAfine-tuning
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

3D Gaussian Splatting (3DGS) renders scenes quickly, but the compression needed for real deployment introduces its own visual damage—floating artifacts, scattered surfaces, and distortions that hit geometry and color separately—which ordinary image-quality metrics miss. The paper builds a 15,200-image benchmark, 3DGS-IEval-15K+, from 10 scenes compressed by 6 algorithms at systematic distortion levels, with human scores for overall, geometry, and color quality (45,600 MOS labels). It then trains 3DGSI-Assessor, a large multimodal model that looks at a rendered image beside its reference and returns all three quality scores in one forward pass. On this benchmark it reports the best correlation with human judgment among more than 30 compared metrics across all three dimensions—SRCC 0.9630 overall, 0.9489 geometry, and 0.9601 color—and it transfers to other novel-view-synthesis benchmarks without retraining. If the results hold, the metric turns quality assessment into a diagnostic: separate geometry and color scores can tell compression engineers which parameter to re-tune.

What carries the argument

The load-bearing mechanism is the hierarchical visual encoder plus two-stage training. Global semantic features come from a frozen pretrained ViT; two Local Dimension-Specific Feature Extraction Modules (LDS-FEM)—ResNet50-based multi-scale extractors pretrained separately to regress geometry and color MOS—capture low-level, dimension-specific distortions. Their outputs are concatenated, projected into the text-embedding space, and fed into an InternVL2.5-8B LMM that first learns to state quality levels as text (stage 1, with LoRA) and then regresses the three MOS values through three MLP heads (stage 2). This is what allows one model to output overall, geometry, and color quality in a single

What would settle it

Train 3DGSI-Assessor on nine scenes and test on the held-out tenth, and also render test views from 3DGS models whose compression parameters never appear in training. If SRCC drops well below the reported within-scene 0.9630, or if a scene-disjoint deep-learning baseline closes the gap with the LMM method, then the SOTA numbers reflect within-scene leakage rather than general 3DGS quality understanding. A second check: compress the same scene twice with identical settings but different random seeds; if predicted scores shift with model identity, the metric is memorizing generators, not just di

Watch

Extended reading notes

Core claim

The paper claims that 3DGS quality assessment must be both representation-specific and multi-dimensional, and that a single large multimodal model can deliver both. It shows that geometry and color quality are only moderately correlated in its dataset (SRCC 0.6505), so an overall score alone cannot say which attribute a compression method damaged. Its method, 3DGSI-Assessor, concatenates global semantic features from a pretrained ViT with local features from two ResNet50-based encoders pretrained to regress geometry and color quality, fuses them inside an LMM, and predicts overall, geometry, and color MOS simultaneously from the LMM's hidden states. In the paper's within-scene evaluation, th

Load-bearing premise

The load-bearing premise is that a 4:1 image-level split within each of only 10 scenes—where training and testing images share the same scene content and the same compressed 3DGS model identities—is a valid test of generalization; if the model memorizes scene- or model-specific distortion patterns, the reported state-of-the-art correlations will not survive contact with unseen scenes.

Editorial extensions

If this is right

  • One forward pass returns overall, geometry, and color quality scores, so compression tuning no longer needs three separately trained per-dimension IQA models.
  • On 3DGS-IEval-15K+, 3DGSI-Assessor reaches SRCC 0.9630 overall, 0.9489 geometry, and 0.9601 color, beating every compared metric and improving on the strongest baseline by 10.3% on geometry and 3.6% on color.
  • Zero-shot evaluation on ENeRF-QA, NeRF-VSQA, and GSC-QA ranks 3DGSI-Assessor first under every correlation measure, suggesting the learned distortion patterns carry across NeRF and 3DGS content.
  • The weak correlations of PSNR, SSIM, and LPIPS with human MOS on this dataset support moving 3DGS compression benchmarking away from pixel-wise fidelity toward learned perceptual metrics.
  • The moderate geometry–color MOS correlation (SRCC 0.65) confirms the two dimensions are separable enough that dimension-specific feedback is meaningful for rate-distortion optimization.

Reading between the lines

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

  • A scene-disjoint split is the natural next test: train on nine scenes and test on the tenth. If 3DGSI-Assessor's margin over per-dimension baselines narrows substantially, part of the reported advantage is memorizing scene- or model-specific compression signatures rather than generic 3DGS distortions.
  • The same architecture could be inverted into a training reward: because it emits separate geometry and color scores, a compression method could optimize its geometry and color parameters against these scores directly, the rate-distortion use the paper lists as future work.
  • The viewpoint-selection criteria (max distance, sparsity, extrapolation, directional novelty) should be re-tested under sparse-view training, where the paper itself notes the train-view versus novel-view quality gap grows; that regime would stress the metric's sensitivity to viewpoint-dependent degradation.
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 introduces 3DGS-IEval-15K+, a large-scale IQA dataset for compressed 3D Gaussian Splatting, containing 15,200 rendered images from 10 scenes, 6 compression algorithms, and 45,600 MOS scores across overall, geometry, and color quality. It also proposes 3DGSI-Assessor, an LMM-based full-reference metric that combines global semantic features with dimension-specific local features and predicts all three quality dimensions in one forward pass. The central claim is state-of-the-art performance on the new benchmark (Table V: SRCC 0.9630 overall, 0.9489 geometry, 0.9601 color) and competitive zero-shot transfer to three external NVS datasets (Table VI). The dataset and code are promised for release.

Significance. If the performance claim survives rigorous evaluation, the paper would make a substantial contribution: it is by far the largest 3DGS IQA dataset, the first with separate geometry/color MOS annotations, and it provides an extensive comparison of handcrafted, deep-learning, and LMM-based metrics. The proposed all-in-one architecture is reasonable and the two-stage LoRA fine-tuning strategy is clearly described. The external zero-shot results on ENeRF-QA, NeRF-VSQA, and GSC-QA are useful independent evidence of transferability and are a genuine strength. However, the headline in-domain SOTA claim rests on an evaluation split that is not group-disjoint and may be inflated by scene/model memorization. The dataset contribution is potentially significant, but the current evidence for the method's generalization on 3DGS-IEval-15K+ is not yet convincing.

major comments (3)
  1. [V-A and Table V] The central SOTA claim is evaluated with a within-scene, image-level 4:1 split: 'for each scene, the rendered images are partitioned 4:1 into training and testing sets.' Because each of the 760 3DGS models is rendered from 20 viewpoints, train and test images share the same scene content and the same model identity, including its unique compression artifacts. A model can therefore memorize scene-specific textures or model-specific distortion patterns instead of learning a generalizable quality function. The 5-fold averaging only changes which views are held out and does not remove this group leakage. The paper provides no scene-disjoint or model-disjoint evaluation. The zero-shot Table VI results partially mitigate the concern, but they do not validate the Table V in-domain claim. Please report performance with a scene-disjoint split (or at least a split where no 3DGS model contributes b
  2. [III-D and III-E] The MOS reliability is not demonstrated. The paper reports 15 annotators per image and Z-score normalization, but no inter-rater agreement statistics (e.g., ICC or Krippendorff's alpha), no outlier subject removal, and no confidence intervals for MOS. Since the entire benchmark and the training labels depend on these MOS values, the lack of reliability evidence weakens the dataset claim. Please add per-dimension inter-rater agreement, subject screening, and an analysis of MOS consistency.
  3. [III-C and Tables II-III] The 'compression level' design is not validated by any bitrate or storage-size measurement. The paper identifies the task as compression quality assessment, but Tables II and III report only parameter settings (e.g., codebook size, prune percentage, lambda), not the resulting file sizes, bits-per-Gaussian, or rate-distortion points. Without this information, it is unclear whether the 16 DL and 6 DL configurations span a monotonic and meaningful compression-quality range. Please report per-model storage sizes and, ideally, rate-distortion curves for each method and level.
minor comments (5)
  1. [III-D, Eq. (6)] The rescaling z'_ij = 100(z_ij + 3)/6 maps z=-3 to 0 and z=3 to 100, but z-scores outside [-3,3] will produce MOS values outside [0,100]. Clarify whether clipping is applied or whether the reported MOS ranges are all within bounds.
  2. [Table II] The table formatting is garbled; the columns for LightGS/C3dGS/Compact-3DGS/CompGS and the geometry/color parameter entries are difficult to parse. Please restructure the table with explicit parameter names and values.
  3. [V-A] The statement that the method 'has been adopted as a baseline in the IEEE Standards Association project 3366.5' is not substantiated by a citation or document. Either provide a reference or remove this non-scientific claim.
  4. [General] There are typos such as 'per iamge' in Section III-D, and the reference format for TReS is inconsistent. A final proofread is recommended.
  5. [I and Table I] The novelty relative to the preliminary conference version [16] is acknowledged only briefly. Please state explicitly in the contribution list which components are new (e.g., geometry/color annotations, the LDS-FEM, the regression heads) to disambiguate from the conference paper.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the core derivation is a standard dataset-plus-supervised-training pipeline, and the top-line in-domain result is cross-checked by zero-shot evaluation on external benchmarks.

full rationale

The paper's derivation chain is self-contained: it builds a subjective dataset from 3DGS renderings and MOS annotations, trains an LMM-based regressor on a held-out split of that dataset, and reports correlations on the test split (Table V) plus zero-shot transfer to external benchmarks (Table VI). No equation or fitted parameter is defined in terms of the predicted quantity, and no step reduces by construction. The within-scene 4:1 image-level split (Section V-A) does create a risk of model-identity leakage between train and test, which could inflate the in-domain SOTA numbers; however, that is a generalization/experimental-design concern, not a circularity in the sense of the prediction being equivalent to the input. The LDS-FEM pretraining uses MOS labels from the same dataset, but the split is not specified; this is an ambiguity worth noting, but there is no quoted evidence that test labels are used in pretraining. Self-citations appear in the related-work and cross-dataset sections (ENeRF-QA, GSC-QA, 3DGS-VBench, 3DGS-IEval-15K), but they are used as background or as external evaluation sets, not as load-bearing justifications for the method's form. The zero-shot results on ENeRF-QA, NeRF-VSQA, and GSC-QA provide independent grounding and are consistent with genuine transfer. Overall, the central claim does not reduce to its inputs.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The work introduces no new physical or conceptual entities; LDS-FEM is an architectural module, not an invented entity. Free parameters are hand-set hyperparameters. The axioms are the domain assumptions about MOS validity, dimension separability, and the evaluation split that the central claims depend on.

free parameters (4)
  • viewpoint selection weights (w_d, w_s, w_e, w_theta) = 0.25 each
    Set by hand in Eq. 1; changing them changes which novel viewpoints are selected for subjective testing.
  • beta (viewpoint feature balance) = 0.3
    Set by hand in Section III-B to balance position and direction; affects training-viewpoint selection.
  • LoRA ranks = r=16 for ViT and LLM
    Chosen for parameter-efficient fine-tuning; final model uses r=16 (Table VII row 8).
  • loss weights lambda_1, lambda_2, lambda_3 = 1, 1, 1
    Set to equal weights in Eq. 13; authors note they are adjustable per application.
assumptions (5)
  • domain assumption MOS averaged from z-scored ratings of 15 subjects per image is a valid ground-truth quality signal
    Section III-D; no inter-rater reliability statistics (ICC, etc.) are reported to validate the labels.
  • domain assumption Overall, geometry, and color quality are separable perceptual dimensions that human raters can judge independently
    Section III-D defines three rating tasks; the inter-dimension correlations in Table IV are moderate, but construct validity is not independently established.
  • ad hoc to paper The within-scene 4:1 image split prevents train/test leakage
    Section V-A: images are split per scene, so all 10 scenes appear in both train and test; this assumes no scene-specific or model-specific memorization affects the reported SOTA.
  • domain assumption ITU-T P.910 11-point impairment scale with side-by-side reference is appropriate for 3DGS image quality
    Used in Section III-D; standard for video, transferred here to still images.
  • domain assumption InternVL2.5-8B and ResNet50 pretrained weights are valid initializations for perceptual QA fine-tuning
    Method Section IV relies on pretrained backbones; no analysis of backbone choice sensitivity.

how reviews work

0 comments
Cite this review

Pith. "Pith review of 3DGSI-Assessor: A Large-Scale Dataset and An LMM-based Method for 3D Gaussian Splatting Image Quality Assessment." pith.science (2026). https://pith.science/paper/34DOGJA5

@misc{pith2026260803279,
  author       = {Pith},
  title        = {Pith review of: 3DGSI-Assessor: A Large-Scale Dataset and An LMM-based Method for 3D Gaussian Splatting Image Quality Assessment},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/34DOGJA5}},
  note         = {Machine review of arXiv:2608.03279}
}
read the original abstract

3D Gaussian Splatting (3DGS) has become a dominant representation for real-time novel view synthesis (NVS), yet its storage footprint makes compression indispensable for practical deployment. 3DGS training and compression introduce representation-specific distortions such as floating artifacts and surface scattering, which conventional image quality assessment (IQA) metrics fail to capture. Moreover, the independent compression of geometric and color attributes may lead to decoupled dimension-specific distortions that must be diagnosed separately, yet existing metrics report only a single overall score. To address these gaps, we present 3DGS-IEval-15K+, a large-scale, multi-dimensional IQA dataset for compressed 3DGS, comprising 15,200 images from 10 diverse scenes, produced by 6 representative 3DGS algorithms at systematically designed compression levels and rendered from 20 strategically selected viewpoints spanning both training views and challenging novel views, annotated with 45,600 mean opinion scores (MOSs) across overall, geometry, and color quality. Based on 3DGS-IEval-15K+, we propose 3DGSI-Assessor, an all-in-one 3DGS IQA framework that integrates global semantic and dimension-specific local features within a large multimodal model (LMM), predicting all three dimensions in a single forward pass. 3DGSI-Assessor achieves state-of-the-art performance on 3DGS-IEval-15K+, and exhibits competitive generalization on other NVS benchmarks. Dataset and code will be released at https://github.com/YukeXing/3DGSI-Assessor.

Figures

Figures reproduced from arXiv: 2608.03279 by the authors.

Figure 1
Figure 1. Illustration of multi-dimensional quality assessment for generative [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of the construction pipeline of 3DGS-IEval-15K+, designed for benchmarking 3DGS representations. (a) We select 6 mainstream compressed [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. 10 selected source content in 3DGS-IEval-15K+: scenes 1-5 depict outdoor scenes, while scenes 6-10 depict indoor scenes. [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Illustration of the proposed viewpoints selection strategy: training viewpoints are selected through feature-based [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Statistical analysis of 3DGS-IEval-15K+ dataset from content complexity and quality perspectives. (a) Scene distribution across colorfulness and [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Framework of 3DGSI-Assessor, an LMM-based method for all-in-one 3DGS image quality assessment. (a) Main functions of 3DGSI-Assessor. The [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

60 extracted references · 45 canonical work pages

  1. [1]

    Nerf: representing scenes as neural radiance fields for view synthesis,

    B. Mildenhall, P. P. Srinivasan, M. Tanciket al., “Nerf: representing scenes as neural radiance fields for view synthesis,”Communications of the ACM (CACM), vol. 65, no. 1, p. 99–106, 2021

    work page 2021

  2. [2]

    3d gaussian splatting for real-time radiance field rendering,

    B. Kerbl, G. Kopanas, T. Leimk ¨uhler, and G. Drettakis, “3d gaussian splatting for real-time radiance field rendering,”ACM Transactions on Graphics (ToG), vol. 42, no. 4, 2023

    work page 2023

  3. [3]

    Scaffold-gs: Structured 3d gaussians for view-adaptive rendering,

    T. Lu, M. Yu, L. Xuet al., “Scaffold-gs: Structured 3d gaussians for view-adaptive rendering,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024, pp. 20 654– 20 664

    work page 2024

  4. [4]

    Hac: Hash-grid assisted context for 3d gaussian splatting compression,

    Y . Chen, Q. Wu, W. Linet al., “Hac: Hash-grid assisted context for 3d gaussian splatting compression,” inEuropean Conference on Computer Vision (ECCV). Springer, 2024, pp. 422–438

    work page 2024

  5. [5]

    Lightgaussian: Unbounded 3d gaussian compression with 15x reduction and 200+ fps,

    Z. Fanet al., “Lightgaussian: Unbounded 3d gaussian compression with 15x reduction and 200+ fps,”Advances in neural information processing systems (NIPS), vol. 37, pp. 140 138–140 158, 2024

    work page 2024

  6. [6]

    Compgs: Smaller and faster gaussian splatting with vector quantization,

    K. Navaneet, K. Pourahmadi Meibodi, S. Abbasi Koohpayegani, and H. Pirsiavash, “Compgs: Smaller and faster gaussian splatting with vector quantization,” inEuropean Conference on Computer Vision (ECCV). Springer, 2024, pp. 330–349

    work page 2024

  7. [7]

    Compressed 3d gaussian splatting for accelerated novel view synthesis,

    S. Niedermayr, J. Stumpfegger, and R. Westermann, “Compressed 3d gaussian splatting for accelerated novel view synthesis,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion (CVPR), June 2024, pp. 10 349–10 358

    2024

  8. [8]

    Compact 3d gaussian splatting for static and dynamic radiance fields,

    J. C. Lee, D. Rhoet al., “Compact 3d gaussian splatting for static and dynamic radiance fields,”arXiv preprint arXiv:2408.03822, 2024

    arXiv 2024

Show all 60 references
  1. [9]

    Eagles: Efficient accelerated 3d gaussians with lightweight encodings,

    S. Girish, K. Gupta, and A. Shrivastava, “Eagles: Efficient accelerated 3d gaussians with lightweight encodings,” inEuropean Conference on Computer Vision (ECCV). Springer, 2024, pp. 54–71

    2024

  2. [10]

    V olseggs: Segmentation and tracking in dynamic volumetric scenes via deformable 3d gaussians,

    S. Yao and C. Wang, “V olseggs: Segmentation and tracking in dynamic volumetric scenes via deformable 3d gaussians,”IEEE Transactions on Visualization and Computer Graphics (TVCG), vol. 32, no. 1, pp. 407– 417, 2026

    2026

  3. [11]

    Explicit-nerf-qa: A quality assessment database for explicit nerf model compression,

    Y . Xing, Q. Yang, K. Yanget al., “Explicit-nerf-qa: A quality assessment database for explicit nerf model compression,” in2024 IEEE Inter- national Conference on Visual Communications and Image Processing (VCIP), 2024, pp. 1–5

    2024

  4. [12]

    Nvs-sqa: Exploring self-supervised quality representation learning for neurally synthesized scenes without references,

    Q. Qu, Y . Shen, X. Chenet al., “Nvs-sqa: Exploring self-supervised quality representation learning for neurally synthesized scenes without references,”arXiv preprint arXiv:2501.06488, 2025

    2025

  5. [13]

    3dgs-vbench: A comprehen- sive video quality evaluation benchmark for 3dgs compression,

    Y . Xing, W. Gordon, Q. Yanget al., “3dgs-vbench: A comprehen- sive video quality evaluation benchmark for 3dgs compression,”arXiv preprint arXiv:2508.07038, 2025

    2025 arXiv

  6. [14]

    Quantifying and alleviating co-adaptation in sparse-view 3d gaussian splatting,

    K. Chen, Y . Zhong, Z. Li, J. Lin, Y . Chen, M. Qin, and H. Wang, “Quantifying and alleviating co-adaptation in sparse-view 3d gaussian splatting,”arXiv preprint arXiv:2508.12720, 2025

    2025

  7. [15]

    Image quality assessment: from error visibility to structural similarity,

    Z. Wang, A. C. Bovik, H. R. Sheikhet al., “Image quality assessment: from error visibility to structural similarity,”IEEE transactions on image processing (TIP), vol. 13, no. 4, pp. 600–612, 2004

    2004

  8. [16]

    3dgs-ieval-15k: a large-scale image quality evaluation database for 3d gaussian-splatting,

    Y . Xing, J. Wang, P. Niuet al., “3dgs-ieval-15k: a large-scale image quality evaluation database for 3d gaussian-splatting,” inProceedings of the 33rd ACM International Conference on Multimedia (ACMMM), 2025, pp. 12 682–12 689. 12

    2025

  9. [17]

    Nerf view synthesis: Subjective quality assessment and objective metrics evaluation,

    P. Martin, A. Rodrigues, J. Ascensoet al., “Nerf view synthesis: Subjective quality assessment and objective metrics evaluation,”IEEE Access, vol. 13, pp. 26–41, 2025

    2025

  10. [18]

    A benchmark for gaussian splatting compression and quality assessment study,

    Q. Yang, K. Yang, Y . Xing, Y . Xu, and Z. Li, “A benchmark for gaussian splatting compression and quality assessment study,” inProceedings of the 6th ACM International Conference on Multimedia in Asia (ACMMM Asia), 2024, pp. 1–8

    2024

  11. [19]

    Nerf-qa: Neural radiance fields quality assessment database,

    P. Martin, A. Rodrigueset al., “Nerf-qa: Neural radiance fields quality assessment database,” in2023 15th International Conference on Quality of Multimedia Experience (QoMEX), 2023, pp. 107–110

    2023

  12. [20]

    Perceptual quality assessment of nerf and neural view synthesis methods for front-facing views,

    H. Liang, T. Wuet al., “Perceptual quality assessment of nerf and neural view synthesis methods for front-facing views,” inComputer Graphics Forum (CGF), vol. 43, no. 2. Wiley Online Library, 2024, p. e15036

    2024

  13. [21]

    Gs-qa: Comprehensive quality assessment benchmark for gaussian splatting view synthesis,

    P. Martin, A. Rodrigues, J. Ascenso, and M. P. Queluz, “Gs-qa: Comprehensive quality assessment benchmark for gaussian splatting view synthesis,” in2025 17th International Conference on Quality of Multimedia Experience (QoMEX). IEEE, 2025, pp. 1–7

    2025

  14. [22]

    Evaluating human perception of novel view synthesis: Subjective quality assessment of gaussian splatting and nerf in dynamic scenes,

    Y . Zhang, J. Maraval, Z. Zhanget al., “Evaluating human perception of novel view synthesis: Subjective quality assessment of gaussian splatting and nerf in dynamic scenes,”arXiv preprint arXiv:2501.08072, 2025

    2025 arXiv

  15. [23]

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

    Z. Chen, W. Wang, Y . Cao, Y . Liu, Z. Gao, E. Cuiet al., “Expanding performance boundaries of open-source multimodal models with model, data, and test-time scaling,”arXiv preprint arXiv:2412.05271, 2024

    2024 arXiv

  16. [24]

    Lora: Low-rank adaptation of large language models

    E. J. Hu, Y . Shen, P. Walliset al., “Lora: Low-rank adaptation of large language models.”International Conference on Learning Representa- tions (ICLR), vol. 1, no. 2, p. 3, 2022

    2022

  17. [25]

    Visual instruction tuning,

    H. Liu, C. Li, Q. Wu, and Y . J. Lee, “Visual instruction tuning,” Advances in neural information processing systems (NIPS), vol. 36, pp. 34 892–34 916, 2023

    2023

  18. [26]

    The unreasonable effectiveness of deep features as a perceptual metric,

    R. Zhang, P. Isolaet al., “The unreasonable effectiveness of deep features as a perceptual metric,” inProceedings of the IEEE conference on computer vision and pattern recognition (CVPR), 2018, pp. 586–595

    2018

  19. [27]

    Improved baselines with visual instruction tuning,

    H. Liu, C. Liet al., “Improved baselines with visual instruction tuning,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR), 2024, pp. 26 296–26 306

    2024

  20. [28]

    Llava-onevision: Easy visual task transfer,

    B. Li, Y . Zhang, D. Guoet al., “Llava-onevision: Easy visual task transfer,”arXiv preprint arXiv:2408.03326, 2024

    2024 arXiv

  21. [29]

    Llava-next-interleave: Tackling multi- image, video, and 3d in large multimodal models,

    F. Li, R. Zhang, H. Zhanget al., “Llava-next-interleave: Tackling multi- image, video, and 3d in large multimodal models,”arXiv preprint arXiv:2407.07895, 2024

    2024 arXiv

  22. [30]

    Deepseek-vl: towards real-world vision- language understanding,

    H. Lu, W. Liu, B. Zhanget al., “Deepseek-vl: towards real-world vision- language understanding,”arXiv preprint arXiv:2403.05525, 2024

    2024 arXiv

  23. [31]

    Qwen2.5-vl technical report,

    S. Bai, K. Chen, X. Liu, J. Wang, W. Ge, S. Songet al., “Qwen2.5-vl technical report,” 2025. [Online]. Available: https: //arxiv.org/abs/2502.13923

    2025 arXiv

  24. [32]

    Learning generalized spatial-temporal deep feature representation for no-reference video quality assessment,

    B. Chen, L. Zhu, G. Li, F. Lu, H. Fan, and S. Wang, “Learning generalized spatial-temporal deep feature representation for no-reference video quality assessment,”IEEE Transactions on Circuits and Systems for Video Technology (TCSVT), vol. 32, no. 4, pp. 1903–1916, 2021

    1903

  25. [33]

    Blind image quality assessment using a deep bilinear convolutional neural network,

    W. Zhang, K. Ma, J. Yanet al., “Blind image quality assessment using a deep bilinear convolutional neural network,”IEEE Transactions on Circuits and Systems for Video Technology (TCSVT), vol. 30, no. 1, p. 36–47, Jan. 2020

    2020

  26. [34]

    Blindly assess image quality in the wild guided by a self-adaptive hyper network,

    S. Su, Q. Yan, Y . Zhu, C. Zhang, X. Ge, J. Sun, and Y . Zhang, “Blindly assess image quality in the wild guided by a self-adaptive hyper network,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR), 2020, pp. 3667–3676

    2020

  27. [35]

    Musiq: Multi- scale image quality transformer,

    J. Ke, Q. Wang, Y . Wang, P. Milanfar, and F. Yang, “Musiq: Multi- scale image quality transformer,” inProceedings of the IEEE/CVF international conference on computer vision (ICCV), 2021, pp. 5148– 5157

    2021

  28. [36]

    Pqa-net: Deep no reference point cloud quality assessment via multi- view projection,

    Q. Liu, H. Yuan, H. Su, H. Liu, Y . Wang, H. Yang, and J. Hou, “Pqa-net: Deep no reference point cloud quality assessment via multi- view projection,”IEEE transactions on circuits and systems for video technology (TCSVT), vol. 31, no. 12, pp. 4645–4660, 2021

    2021

  29. [37]

    Nerf-nqa: No-reference quality assessment for scenes generated by nerf and neural view synthesis methods,

    Q. Qu, H. Liang, X. Chen, Y . Y . Chung, and Y . Shen, “Nerf-nqa: No-reference quality assessment for scenes generated by nerf and neural view synthesis methods,”IEEE Transactions on Visualization and Computer Graphics (TVCG), vol. 30, no. 5, pp. 2129–2139, 2024

    2024

  30. [38]

    Mip-nerf 360: Unbounded anti-aliased neural radiance fields,

    J. T. Barron, B. Mildenhallet al., “Mip-nerf 360: Unbounded anti-aliased neural radiance fields,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR), 2022, pp. 5470–5479

    2022

  31. [39]

    Tanks and temples: Benchmarking large-scale scene reconstruction,

    A. Knapitsch, J. Park, Q.-Y . Zhou, and V . Koltun, “Tanks and temples: Benchmarking large-scale scene reconstruction,”ACM Transactions on Graphics (ToG), vol. 36, no. 4, pp. 1–13, 2017

    2017

  32. [40]

    Deep blending for free-viewpoint image-based rendering,

    P. Hedman, J. Philip, T. Priceet al., “Deep blending for free-viewpoint image-based rendering,”ACM Transactions on Graphics (ToG), vol. 37, no. 6, pp. 1–15, 2018

    2018

  33. [41]

    Subjective video quality assessment methods for multimedia applications,

    P. ITU-T RECOMMENDATION, “Subjective video quality assessment methods for multimedia applications,” 1999

    1999

  34. [42]

    500-14, methodologies for the subjective assessment of the quality of television images,

    I. R. BT, “500-14, methodologies for the subjective assessment of the quality of television images,”Geneva: International Telecommunication Union, 2019

    2019

  35. [43]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” inProceedings of the IEEE conference on computer vision and pattern recognition (CVPR), 2016, pp. 770–778

    2016

  36. [44]

    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

    2012

  37. [45]

    Multiscale structural similarity for image quality assessment,

    Z. Wang, E. P. Simoncelli, and A. C. Bovik, “Multiscale structural similarity for image quality assessment,” inThe Thrity-Seventh Asilomar Conference on Signals, Systems & Computers, 2003, vol. 2. Ieee, 2003, pp. 1398–1402

    2003

  38. [46]

    Information content weighting for perceptual image quality assessment,

    Z. Wang and Q. Li, “Information content weighting for perceptual image quality assessment,”IEEE Transactions on image processing (TIP), vol. 20, no. 5, pp. 1185–1198, 2010

    2010

  39. [47]

    Image information and visual quality,

    H. R. Sheikh and A. C. Bovik, “Image information and visual quality,” IEEE Transactions on image processing (TIP), vol. 15, no. 2, pp. 430– 444, 2006

    2006

  40. [48]

    Fsim: A feature similarity index for image quality assessment,

    L. Zhang, L. Zhang, X. Mou, and D. Zhang, “Fsim: A feature similarity index for image quality assessment,”IEEE transactions on Image Processing (TIP), vol. 20, no. 8, pp. 2378–2386, 2011

    2011

  41. [49]

    No-reference image quality assessment in the spatial domain,

    A. Mittal, A. K. Moorthy, and A. C. Bovik, “No-reference image quality assessment in the spatial domain,”IEEE Transactions on image processing (TIP), vol. 21, no. 12, pp. 4695–4708, 2012

    2012

  42. [50]

    Image quality assessment: Unifying structure and texture similarity,

    K. Ding, K. Ma, S. Wang, and E. P. Simoncelli, “Image quality assessment: Unifying structure and texture similarity,”IEEE transactions on pattern analysis and machine intelligence (TPAMI), vol. 44, no. 5, pp. 2567–2581, 2020

    2020

  43. [51]

    mplug-owl3: Towards long image-sequence understanding in multi-modal large language models,

    J. Ye, H. Xu, H. Liuet al., “mplug-owl3: Towards long image-sequence understanding in multi-modal large language models,”arXiv preprint arXiv:2408.04840, 2024

    2024 arXiv

  44. [52]

    Exploring clip for assessing the look and feel of images,

    J. Wang, K. C. Chan, and C. C. Loy, “Exploring clip for assessing the look and feel of images,” inProceedings of the AAAI conference on artificial intelligence (AAAI), vol. 37, no. 2, 2023, pp. 2555–2563

    2023

  45. [53]

    Cogagent: A visual language model for gui agents,

    W. Hong, W. Wang, Q. Lvet al., “Cogagent: A visual language model for gui agents,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR), 2024, pp. 14 281–14 290

    2024

  46. [54]

    Gemini 1.5: Unlocking mul- timodal understanding across millions of tokens of context,

    G. Team, P. Georgiev, V . I. Leiet al., “Gemini 1.5: Unlocking mul- timodal understanding across millions of tokens of context,”arXiv preprint arXiv:2403.05530, 2024

    2024 arXiv

  47. [55]

    Q-align: Teaching lmms for visual scoring via discrete text-defined levels,

    H. Wu, Z. Zhang, W. Zhanget al., “Q-align: Teaching lmms for visual scoring via discrete text-defined levels,” 2023

    2023

  48. [56]

    LLaMA 3.2: Revolutionizing Edge AI and Vision with Open, Customizable Models,

    Meta, “LLaMA 3.2: Revolutionizing Edge AI and Vision with Open, Customizable Models,” https://ai.meta.com/blog/ llama-3-2-revolutionizing-edge-ai-and-vision/, 2024, meta AI Blog, retrieved December 2024

    2024

  49. [57]

    No-reference im- age quality assessment via transformers, relative ranking, and self- consistency,

    S. A. Golestaneh, S. Dadsetan, and K. M. Kitani, “No-reference im- age quality assessment via transformers, relative ranking, and self- consistency,” inProceedings of the IEEE/CVF winter conference on applications of computer vision (CVPR), 2022, pp. 1220–1230

    2022

  50. [58]

    Maniqa: Multi-dimension attention network for no-reference image quality assessment,

    S. Yang, T. Wu, S. Shiet al., “Maniqa: Multi-dimension attention network for no-reference image quality assessment,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR), 2022, pp. 1191–1200

    2022

  51. [59]

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

    W. Sun, X. Min, D. Tuet al., “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, p. 1178–1192, Nov. 2023

    2023

  52. [60]

    Blind image quality assessment via vision-language correspondence: A multitask learning perspective,

    W. Zhang, G. Zhai, Y . Wei, X. Yang, and K. Ma, “Blind image quality assessment via vision-language correspondence: A multitask learning perspective,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR), 2023, pp. 14 071–14 081

    2023

Pith tools

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