Pith. sign in

REVIEW 3 major objections 5 minor 117 references

F-Bench: Rethinking Human Preference Evaluation Metrics for Benchmarking Face Generation, Customization, and Restoration

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper introduces FaceQ, a large-scale database of human preference ratings for AI-generated, customized, and restored faces, and F-Eval, a single evaluator that outperforms existing quality metrics on all four dimensions.

desk verdict FaceQ is a genuinely useful new benchmark resource, but F-Eval's 'optimal across all tasks' claim rests on an unexplained 80:20 split with likely group-level leakage, so the SOTA numbers need revision before they can be trusted. read the letter →

arxiv 2412.13155 v2 pith:6AJ3IP2X submitted 2024-12-17 cs.CV

Lu Liu , Huiyu Duan , Qiang Hu , Liu Yang , Chunlei Cai , Tianxiao Ye , Huayu Liu , Xiaoyun Zhang
show 1 more author
Guangtao Zhai
This is my paper · ORCID
classification cs.CV
keywords AI-generatedfaceimagesgenerationcustomizationrestorationimagequalityassessmenthumanpreferencebenchmarklargemultimodalmodel
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

The paper sets out to establish that human preferences for AI-generated, customized, and restored faces are multidimensional and are not captured by existing image-quality metrics. It introduces FaceQ, a database of 12,255 face images produced by 29 models, with 491,130 human ratings converted into 32,742 mean opinion scores across four dimensions: quality, authenticity, identity fidelity, and text-image correspondence. On this data it benchmarks 26 existing IQA, face IQA, and preference metrics and finds that standard metrics correlate weakly with human judgment on authenticity, identity fidelity, and correspondence. It then builds F-Eval, a large-multimodal-model evaluator with a frozen face encoder and dimension-specific adapters, and reports that F-Eval outperforms all baselines on all three tasks and all four dimensions. If this holds, the paper supplies the first large-scale benchmark and one-for-all evaluator for how well generative face models match human preferences.

What carries the argument

The load-bearing object is FaceQ, whose subjective scores come from 180 annotators rating images under ITU-R BT.500-style protocols, with kurtosis-based outlier rejection and Z-score rescaling to a 0-100 MOS range. The four rating dimensions define the evaluation space: quality is overall perceptual fidelity, authenticity is how much the image resembles a real photograph, identity fidelity is how well the reference identity survives customization or restoration, and correspondence is text-image alignment. The evaluator F-Eval is an instruction-tuned large multimodal model, built on Qwen2.5-VL, that accepts an image, optional reference image, and prompt, and returns a score per requested dimension. Its distinctive mechanism is a mixture of LoRA experts: four low-rank adapters, one per dimension, are activated by a router that classifies the user's question, so the same weights serve all tasks without cross-dimension interference. A frozen ArcFace face encoder supplies identity-structure features that the vision encoder alone would miss.

What would settle it

Re-split FaceQ by clustering images on source identity, prompt, and low-quality input so no near-duplicate crosses the split, then retrain and re-evaluate F-Eval; if its SRCC or PLCC drops substantially relative to the reported tables, the claim that F-Eval is the best one-for-all evaluator on new faces is not supported.

Watch

Extended reading notes

Core claim

The central claim is that the perceived quality of a generated face is not a single number: authenticity, identity fidelity, and text-image correspondence move independently, and a usable evaluator must score them separately. FaceQ-Gen covers 14 text-to-face generators, FaceQ-Cus covers six identity-driven customization methods with a reference image, and FaceQ-Res covers nine restoration models on both synthetic and real-world low-quality inputs. Benchmarking on these subsets shows that generation models now follow prompts well but lag in authenticity, customization models that preserve identity often sacrifice quality, and restoration models maintain quality better than identity. F-Eval is designed to predict all four dimensions in one pass by fusing a vision encoder, a frozen face encoder (ArcFace), and a pretrained large language model, with four LoRA experts chosen by a trainable router according to the dimension being asked. The paper reports that F-Eval achieves the best correlation with human scores across every task and dimension, with the largest gains on customization, correspondence, and identity fidelity.

Load-bearing premise

F-Eval's reported accuracy assumes the 80/20 training-test split of FaceQ keeps near-duplicate views, repeated identities, and the same prompts or restoration sources out of both sides; if that separation leaks, the headline correlations overstate performance on genuinely new faces.

Editorial extensions

If this is right

  • FaceQ gives the first standardized way to rank face generation, customization, and restoration models by human preference across separate dimensions rather than by a single quality score.
  • Researchers evaluating generative face models should treat authenticity, identity fidelity, and correspondence as distinct report cards, since models that score high on one can score low on another.
  • Existing IQA and face-IQA methods are not reliable proxies for human preference on AI-generated faces, so benchmark claims based on them need re-examination.
  • F-Eval provides a single instruction-tuned model that can be applied directly to new generated, customized, or restored face images without retraining per dimension.

Reading between the lines

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

  • Editorial extension: if FaceQ's 80/20 split is not deduplicated by identity, prompt, or low-quality source image, the reported superiority of F-Eval could shrink under an identity-aware re-split.
  • Editorial extension: the same four-dimension, multi-expert instruction-tuning design could transfer to evaluating AI-generated full-body images, videos, or avatar faces, where authenticity and identity drift also matter.
  • The paper notes in its supplementary material that F-Eval degrades on very small or extreme side-view faces and does not localize distorted regions; a natural follow-up is a face encoder trained on such poses and a region-level explanation head.
  • Editorial extension: FaceQ's multi-dimensional MOSs could be reused as a preference-aligned reward signal for tuning generation models, though the paper itself does not test that use.
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. This paper introduces FaceQ, a large-scale database of 12,255 AI-generated, customized, and restored face images with 491,130 human ratings aggregated into 32,742 MOS values across four dimensions (quality, authenticity, ID fidelity, correspondence), covering 29 generative models. Using FaceQ, the authors construct F-Bench, a benchmark that (i) compares the 29 models across tasks and prompt categories, and (ii) evaluates 26 existing IQA/FQA/AIGCIQA/preference metrics against human MOS. The paper further proposes F-Eval, a Qwen-VL-based LMM with four dimension-specific LoRA experts and a router, which takes the image(s), prompt text, and optional reference image as input and outputs per-dimension quality scores. Reported results show that existing metrics correlate weakly with human MOS on FaceQ, and that F-Eval attains the highest SRCC/KRCC/PLCC in nearly all task-dimension pairs.

Significance. If the dataset and benchmark are released as described, FaceQ would be the first large-scale face-focused AIGC quality database with multi-dimensional human preference labels, and F-Bench would be a practically useful resource for the face generation, customization, and restoration communities. The observation that existing metrics—including recent AIGC-specific ones such as CLIPScore, ImageReward, and MINTIQA—correlate poorly with human MOS on AI-generated faces is an important cautionary result that appears well supported by the reported correlations, which are derived from independent human ratings and are not circular. The paper also provides substantial methodological detail (annotation interface, ITU-R BT.500-14 procedure, degradation pipelines, model implementation details) that increases the usability of the resource. The proposed F-Eval architecture, combining a frozen ArcFace face encoder with dimension-specific LoRA experts, is a reasonable design for a one-for-all evaluator.

major comments (3)
  1. [Section 6.2 / 6.3] The evaluation protocol for F-Eval is not specified. Section 6.2 states that the deep-learning baselines are "retrained on FaceQ with an 80:20 training/testing split," but Section 6.1, which describes F-Eval's training, does not mention any split and says only that the network is trained for 10 epochs on FaceQ. The reader therefore cannot determine whether F-Eval's test images in Tables 2–4 were held out during training. If F-Eval was trained on the full dataset, its reported SRCC/KRCC/PLCC values reflect memorization, which would invalidate the Section 6.3 claim that "Our method achieves optimal performance across all tasks and dimensions." The authors must state the exact split used for F-Eval and, if it was not the same 80:20 split used for the baselines, re-run the comparison under a shared protocol.
  2. [Section 6.2 / 3.1] The dataset is structured by shared source content, and a random image-level split can leak source-level information. FaceQ-Gen (Section 3.1.1) generates multiple images from a common prompt list across 14 models; FaceQ-Cus (Section 3.1.2) reuses reference identities and a subset of the same prompts; FaceQ-Res (Section 3.1.3) restores the same synthetic and real-world low-quality inputs with nine methods. Under an image-level 80:20 split, images derived from the same prompt, identity, or LQ source can appear in both training and test partitions. Because F-Eval explicitly receives the prompt text and reference image as inputs, it is the method best positioned to exploit such leakage, particularly for the correspondence and ID-fidelity dimensions. The manuscript does not report any group-wise splitting, source-based deduplication, or identity-aware partitioning. The authors should re-evaluate F-Eval and the fine-tuned baselines on a group-wise split (e.g., by prompt, identity, and LQ source) or justify why image-level splitting is adequate given the shared-source structure.
  3. [Table 2 / Section 6.3] The state-of-the-art claim is based on a single 80:20 split with no statistical significance assessment. The reported margins between F-Eval and the best fine-tuned baseline are small in several cells (e.g., FaceQ-Gen Quality SRCC 0.8486 vs. 0.8312; FaceQ-Res synthetic Quality SRCC 0.8692 vs. 0.8656), and no confidence intervals, bootstrap tests, or multiple-seed results are provided. Without variance estimates or significance tests, the claim that F-Eval "achieves optimal performance across all tasks and dimensions" is not supported beyond the specific split shown. The authors should report mean±std over multiple splits or provide significance tests for the key comparisons.
minor comments (5)
  1. [Section 3.3 / Supplementary 9.1] The main text says "outlier detection based on Kurtosis is conducted with a rejection rate of 3% [78]," while the supplementary (Section 9.1) says "3% invalid data are removed and no subject is removed"; these statements should be reconciled to specify whether ratings or subjects were rejected.
  2. [Tables 3 and 4] Tables 3 and 4 cite SER-FIQ as [25], but in the reference list [25] is TReS (Golestaneh et al.); the correct reference for SER-FIQ is [85]. The citation inconsistency should be fixed.
  3. [Abstract / Introduction] The abstract contains a typo: "restorted faces" should be "restored faces." The introduction also uses "state-of-art" where "state-of-the-art" is intended.
  4. [Section 6.2] "The comparison QA models falls into four categories" should be "The comparison QA models fall into four categories."
  5. [Availability] The paper states that the FaceQ database "will be publicly available upon publication," but no data or evaluation code are provided in the current version, which limits independent verification of the reported correlations and benchmark findings; the authors should clarify the release timeline and, if possible, provide evaluation code to enable reproducibility.

Circularity Check

0 steps flagged · score 2.0 of 10

No load-bearing circularity: the human-MOS benchmark and the "existing metrics fail" conclusion are independent; F-Eval is a supervised fit on the same benchmark, and self-citations are not load-bearing.

full rationale

The central empirical claims rest on human mean opinion scores collected independently of any proposed metric: FaceQ is constructed from subjective ratings, and the conclusion that existing IQA/FQA/AIGCIQA metrics are "relatively ineffective" follows from correlating those off-the-shelf or fine-tuned methods with the human MOS. That is an external, non-circular comparison. F-Eval is trained on FaceQ MOS using the losses in Eqs. (13)-(16) and then evaluated on FaceQ; if the evaluation uses the same held-out split applied to the fine-tuned baselines, this is standard supervised benchmark evaluation, not a derivation equivalent to its inputs. The paper states in Sec. 6.2 that the classical deep models are "retrained on FaceQ with an 80:20 training/testing split" but does not explicitly state which split is used for F-Eval, and the group-structured content (shared prompts, reference identities, and LQ inputs) could create leakage that inflates correlations; these are experimental-reporting and validity concerns, not demonstrated circularity. Several references are to the authors' own prior work (e.g., AIGCIQA2023, MINTIQA, FineVQ), but they appear only as related work or baselines and are not used as the justification for FaceQ or F-Eval's design. No self-definitional, fitted-input-as-prediction, or self-citation-chain reduction is exhibited in the paper, so no specific circular step is flagged; the low score reflects minor self-citations and the supervised same-benchmark nature of F-Eval, not load-bearing circularity.

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

The benchmark results are empirical and rest mainly on the reliability of subjective scores and representativeness of the collected images. The proposed F-Eval's superiority depends further on the training hyperparameters and on the train/test split being leakage-free. No invented entities are introduced; the method is built from existing components.

free parameters (6)
  • LoRA ranks per dimension = 16, 32, 16, 16
    Hand-chosen ranks for the quality, authenticity, ID fidelity, and correspondence LoRA experts; not tuned on a validation set, they affect F-Eval's capacity and final scores.
  • LoRA alpha (scaling factor) = Equal to rank
    Set equal to rank per standard LoRA practice; affects the magnitude of the low-rank update.
  • LoRA dropout = 0.05
    Standard regularization choice; not justified by a sweep.
  • Learning rate = 1e-4
    Chosen for AdamW; no learning rate search is reported.
  • Training epochs = 10
    Stopping criterion not described; could influence reported performance.
  • Batch size = 4
    Hardware-constrained choice on two RTX 3090 GPUs.
assumptions (5)
  • domain assumption Human MOS obtained through ITU-R BT.500-14 and z-score normalization are valid, commensurable measures of human preference.
    All benchmark correlations and F-Eval training target these MOS (Section 3.3, Eqs. 1-3).
  • ad hoc to paper The four dimensions (quality, authenticity, ID fidelity, correspondence) adequately decompose human preference for AIGFs.
    The paper selects these dimensions without an independent derivation; the decomposition is a design choice.
  • domain assumption The 29 models, nine prompt categories, and two synthetic degradation pipelines plus real-world sources are representative of the AIGF landscape.
    Used to generalize benchmark conclusions beyond the specific images collected (Sections 3.1 and 4).
  • ad hoc to paper The 80:20 train/test split does not leak near-duplicate identities, prompts, or degraded sources, so correlations reflect generalization.
    No deduplication or identity-aware split is described (Section 6.2); leakage would inflate F-Eval and fine-tuned baseline scores.
  • domain assumption No annotator-level effects remain after z-score normalization and 3% outlier rejection.
    Eqs. 1-3 assume per-annotator normalization removes scale and offset differences; the paper does not check inter-group agreement.

how reviews work

0 comments
Cite this review

Pith. "Pith review of F-Bench: Rethinking Human Preference Evaluation Metrics for Benchmarking Face Generation, Customization, and Restoration." pith.science (2026). https://pith.science/paper/6AJ3IP2X

@misc{pith2026241213155,
  author       = {Pith},
  title        = {Pith review of: F-Bench: Rethinking Human Preference Evaluation Metrics for Benchmarking Face Generation, Customization, and Restoration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6AJ3IP2X}},
  note         = {Machine review of arXiv:2412.13155}
}
read the original abstract

Artificial intelligence generative models exhibit remarkable capabilities in content creation, particularly in face image generation, customization, and restoration. However, current AI-generated faces (AIGFs) often fall short of human preferences due to unique distortions, unrealistic details, and unexpected identity shifts, underscoring the need for a comprehensive quality evaluation framework for AIGFs. To address this need, we introduce FaceQ, a large-scale, comprehensive database of AI-generated Face images with fine-grained Quality annotations reflecting human preferences. The FaceQ database comprises 12,255 images generated by 29 models across three tasks: (1) face generation, (2) face customization, and (3) face restoration. It includes 32,742 mean opinion scores (MOSs) from 180 annotators, assessed across multiple dimensions: quality, authenticity, identity (ID) fidelity, and text-image correspondence. Using the FaceQ database, we establish F-Bench, a benchmark for comparing and evaluating face generation, customization, and restoration models, highlighting strengths and weaknesses across various prompts and evaluation dimensions. Additionally, we assess the performance of existing image quality assessment (IQA), face quality assessment (FQA), AI-generated content image quality assessment (AIGCIQA), and preference evaluation metrics, manifesting that these standard metrics are relatively ineffective in evaluating authenticity, ID fidelity, and text-image correspondence. The FaceQ database will be publicly available upon publication.

Figures

Figures reproduced from arXiv: 2412.13155 by the authors.

Figure 1
Figure 1. Data construction pipeline of F-bench and content overview of FaceQ database. From left to right, (a) Diverse input source data for face generation, customization, and restoration. (b) F-Bench benchmarks 29 face generative models, including 14 face generation models, 6 face customization models, and 9 face restoration models. (c) Multi-dimensional subjective preference evaluation in F-Bench. (d) 3D scatter plots of … view at source ↗
Figure 2
Figure 2. Rating comparisons of eight dimensions. Each column presents a pair of intuitive examples of each dimension, with red indicating the better rating and blue indicating the worse one. From left to right, the subsets are face generation, face customization, and face restoration subsets. The last row displays the corresponding prompts, reference image-prompt pairs, and the GT-LQ image pairs. the biometric utility for fa… view at source ↗
Figure 3
Figure 3. Mean Opinion Score (MOS) distribution histograms and kernel density curves. (a)-(c) MOS distributions for three subsets. (d)-(h) Model-wise MOS distributions. MOS Distributions of other methods are provided in the supplementary material. Authenticity. This dimension evaluates how closely the generated image resembles a natural, real-life photograph, with particular emphasis on realistic skin texture, facial de￾tails… view at source ↗
Figures from the paper (16 more)
Figure 4
Figure 4. Figure 4: Average MOS score comparison across all models and dimensions. (a) Face generation. (b) Face customization. (c) Face restoration. The models are arranged in a clockwise order by release date [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Violin plots of quality, authenticity and correspondance scores in FaceQ-Gen subset across nine prompt categories. Correspondence. Correspondence scores show a steady in￾crease over time, indicating consistent improvement in text comprehension. Quality scores generally…
Figure 6
Figure 6. Figure 6: The overall framework of F-Eval. F-Eval can evaluate quality, authenticity, correspondence, and identity fidelity in a one-for￾all framework. It can process both single and paired images, along with prompts, to produce quality scores. It consists of three encoders, inc…
Figure 7
Figure 7. Figure 7: FaceQ-Gen Examples [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]
Figure 8
Figure 8. Figure 8: FaceQ-Cus Examples. illustrated, the MOS scores effectively and intuitively cap￾ture the strengths and weaknesses of the images, accurately reflecting human preferences across different dimensions. Similarly, [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]
Figure 9
Figure 9. Figure 9: FaceQ-Res Examples. ing performance between real-world and synthetic inputs, resulting in noticeable differences in their distributions, as exemplified by SPARNet [15]. 10.2. Perspectiive Analysis To provide a clearer comparison of the strengths and weak￾nesses of diff…
Figure 10
Figure 10. Figure 10: Comparisons of the selected four low-level feature distributions calculated on proposed FaceQ dataset. 0 0.2 0.4 0.6 0.8 1 FaceQ-Gen FaceQ-Cus FaceQ-Res C o n v e r a g e U nifo r mit y Brightness Contrast Colorfulness Sharpness 0 0.2 0.4 0.6 0.8 1 1.2 FaceQ-Gen FaceQ…
Figure 11
Figure 11. Figure 11: Comparisons of the selected four low-level features calculated on the proposed FaceQ dataset. (a) Coverage uniformity. (b) Relative range. (a) (b) (c) (d) [PITH_FULL_IMAGE:figures/full_fig_p019_11.png]
Figure 12
Figure 12. Figure 12: Screenshots of the rating interface for human evaluation. (a) Face generation evaluation interface. (b) Face customization evaluation interface. (c) Face restoration (real world) evaluation interface. (d) Face restoration (synthetic) evaluation interface [PITH_FULL_I…
Figure 13
Figure 13. Figure 13: Additional visualizations of the 3D scatter of MOSs. We sample five representative points from the scatter and visualize their MOS scores across three dimensions. Each dimension is represented by a different color. The relatively low dimensions are underlined. (a) Fac…
Figure 14
Figure 14. Figure 14: MOS distribution histograms and kernel density curves across different face generation models. “full steps”contains images generated in default sampling steps and “1/4 steps”contains the images generated by one-quarter of the default steps. ELITE FastComposer InstantI…
Figure 16
Figure 16. Figure 16: MOS distribution histograms and kernel density curves across different face restoration models. “synthetic”refers to images restored from the synthetic low-quality inputs while “real world”refers to the images restored from real-world low-quality inputs. (a) (b) (c) …
Figure 17
Figure 17. Figure 17: Comparison of averaged MOS of different models across Quality,Authenticity, ID Fidelity, and Correspondence. (a) Face generation models. (b) Face customization models. (c) Face restoration models. 11. QA Methods Implementation Details 11.1. Evaluation Metrics We adopt…
Figure 18
Figure 18. Figure 18: Comparison of different model rankings based on the averaged MOS (a) Face generation models. (b) Face customization models. (c) Face restoration models. (a) (b) (c) Young Middle-aged Old Young Middle-aged Old Young Middle-aged Old [PITH_FULL_IMAGE:figures/full_fig_p0…
Figure 19
Figure 19. Figure 19: Comparison of multi-dimensional MOS distributions across age groups. “Authen.”, “Corres.”and “ID Fide.”denote Au￾thenticity, Correspondence, and ID Fidelity respectively. (a) Face generation models. (b) Face customization models. (c) Face restoration models. loss func…
Figure 20
Figure 20. Figure 20: Comparison of multi-dimensional MOS distributions across genders. “Authen.”, “Corres.”and “ID Fide.”denote Authentic￾ity, Correspondence, and ID Fidelity respectively. (a) Face generation models. (b) Face customization models. (c) Face restoration models [PITH_FULL_I…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

117 extracted references · 55 canonical work pages

  1. [1]

    Deep Floyd.https://github.com/deep-floyd/ IF, Accessed: 2024-10-03. 2, 3, 1

    2024

  2. [2]

    co / dreamlike - art / dreamlike - photoreal - 2

    Dreamlike V2.https : / / huggingface . co / dreamlike - art / dreamlike - photoreal - 2 . 0, Accessed: 2024-10-03. 5, 1

    2024

  3. [3]

    co / black - forest-labs/FLUX.1-dev, Accessed: 2024-10-03

    Flux-dev.https : / / huggingface . co / black - forest-labs/FLUX.1-dev, Accessed: 2024-10-03. 1, 3, 5, 4

    2024

  4. [4]

    co/h94/IP-Adapter, Accessed: 2024-10-03

    IP-Adapter-FaceID-SDXL.https://huggingface. co/h94/IP-Adapter, Accessed: 2024-10-03. 3, 5, 1

    2024

  5. [5]

    co/h94/IP-Adapter-FaceID, Accessed: 2024-10-

    IP-Adapter-FaceID-Plus.https : / / huggingface . co/h94/IP-Adapter-FaceID, Accessed: 2024-10-

    2024

  6. [6]

    co / Kwai - Kolors/Kolors, Accessed: 2024-10-03

    Kolors.https : / / huggingface . co / Kwai - Kolors/Kolors, Accessed: 2024-10-03. 3, 1, 5

    2024

  7. [7]

    co / stablediffusionapi/protovision- xl- v6.6, Accessed: 2024-10-03

    ProtoVision V6.6.https : / / huggingface . co / stablediffusionapi/protovision- xl- v6.6, Accessed: 2024-10-03. 3, 1

    2024

  8. [8]

    3, 5, 1, 4

    Realistic Vision V5.1.https://huggingface.co/ SG161222/Realistic_Vision_V5.1_noVAE, Ac- cessed: 2024-10-03. 3, 5, 1, 4

    2024

Show all 117 references
  1. [9]

    SD3.https://huggingface.co/stabilityai/ stable-diffusion-3-medium, Accessed: 2024-10-

    2024

  2. [10]

    GPT-4o.https://chatgpt.com/, Accessed: 2024- 10-10. 3

    2024

  3. [11]

    Shuai Bai, Keqin Chen, Xuejing Liu, Jialin Wang, Wenbin Ge, Sibo Song, Kai Dang, Peng Wang, Shijie Wang, Jun Tang, et al. Qwen2. 5-vl technical report.arXiv preprint arXiv:2502.13923, 2025. 2, 6, 7

    2025 arXiv

  4. [12]

    Cr-fiqa: face image quality assessment by learning sample relative classifiability

    Fadi Boutros, Meiling Fang, Marcel Klemt, Biying Fu, and Naser Damer. Cr-fiqa: face image quality assessment by learning sample relative classifiability. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 5836–5845, 2023. 2

    2023

  5. [13]

    An image quality assessment dataset for portraits

    Nicolas Chahine, Stefania Calarasanu, Davide Garcia- Civiero, Th´eo Cayla, Sira Ferradans, and Jean Ponce. An image quality assessment dataset for portraits. InProceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 9968–9978, 2023. 2

    2023

  6. [14]

    Promptiqa: Boosting the performance and gen- eralization for no-reference image quality assessment via prompts

    Chen et al. Promptiqa: Boosting the performance and gen- eralization for no-reference image quality assessment via prompts. InECCV, 2024. 7, 8

    2024

  7. [15]

    Learning spatial attention for face super-resolution.IEEE Transactions on Image Processing (TIP), 30:1219–1231, 2020

    Chaofeng Chen, Dihong Gong, Hao Wang, Zhifeng Li, and Kwan-Yee K Wong. Learning spatial attention for face super-resolution.IEEE Transactions on Image Processing (TIP), 30:1219–1231, 2020. 2, 3, 5, 1

    2020

  8. [16]

    Progressive semantic- aware style transformation for blind face restoration

    Chaofeng Chen, Xiaoming Li, Lingbo Yang, Xianhui Lin, Lei Zhang, and Kwan-Yee K Wong. Progressive semantic- aware style transformation for blind face restoration. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition (CVPR), pages 11896–11905,

  9. [17]

    Pixart-α: Fast train- ing of diffusion transformer for photorealistic text-to-image synthesis.arXiv preprint arxiv:2310.00426, 2023

    Junsong Chen, Jincheng Yu, Chongjian Ge, Lewei Yao, Enze Xie, Yue Wu, Zhongdao Wang, James Kwok, Ping Luo, Huchuan Lu, and Zhenguo Li. Pixart-α: Fast train- ing of diffusion transformer for photorealistic text-to-image synthesis.arXiv preprint arxiv:2310.00426, 2023. 2, 3, 1, 4

    arXiv

  10. [18]

    Dsl-fiqa: Assess- ing facial image quality via dual-set degradation learning and landmark-guided transformer

    Wei-Ting Chen, Gurunandan Krishnan, Qiang Gao, Sy- Yen Kuo, Sizhou Ma, and Jian Wang. Dsl-fiqa: Assess- ing facial image quality via dual-set degradation learning and landmark-guided transformer. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogniti...

    2024

  11. [19]

    Fsrnet: End-to-end learning face super-resolution with facial priors

    Yu Chen, Ying Tai, Xiaoming Liu, Chunhua Shen, and Jian Yang. Fsrnet: End-to-end learning face super-resolution with facial priors. InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR), pages 2492–2501, 2018. 2

    2018

  12. [20]

    Idadapter: Learn- ing mixed features for tuning-free personalization of text- to-image models.arXiv preprint arXiv:2403.13535, 2024

    Siying Cui, Jiankang Deng, Jia Guo, Xiang An, Yongle Zhao, Xinyu Wei, and Ziyong Feng. Idadapter: Learn- ing mixed features for tuning-free personalization of text- to-image models.arXiv preprint arXiv:2403.13535, 2024. 2

    2024 arXiv

  13. [21]

    Flashattention-2: Faster attention with better par- allelism and work partitioning

    Tri Dao. Flashattention-2: Faster attention with better par- allelism and work partitioning. InProceedings of the Inter- national Conference on Learning Representations (ICLR). 7

  14. [22]

    Arcface: Additive angular margin loss for deep face recognition

    Jiankang Deng, Jia Guo, Niannan Xue, and Stefanos Zafeiriou. Arcface: Additive angular margin loss for deep face recognition. InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR), pages 4690–4699, 2019. 6, 7, 8

    2019

  15. [23]

    Finevq: Fine-grained user generated content video quality assessment

    Huiyu Duan, Qiang Hu, Jiarui Wang, Liu Yang, Zitong Xu, Lu Liu, Xiongkuo Min, Chunlei Cai, Tianxiao Ye, Xiaoyun Zhang, et al. Finevq: Fine-grained user generated content video quality assessment. Inm, 2025. 2, 3

    2025

  16. [24]

    An image is worth one word: Personalizing text-to- image generation using textual inversion.arXiv preprint arXiv:2208.01618, 2022

    Rinon Gal, Yuval Alaluf, Yuval Atzmon, Or Patash- nik, Amit H Bermano, Gal Chechik, and Daniel Cohen- Or. An image is worth one word: Personalizing text-to- image generation using textual inversion.arXiv preprint arXiv:2208.01618, 2022. 2

    2022 arXiv

  17. [25]

    Alireza Golestaneh, Saba Dadsetan, and Kris M

    S. Alireza Golestaneh, Saba Dadsetan, and Kris M. Kitani. No-reference image quality assessment via transformers, relative ranking, and self-consistency. InProceedings of the IEEE/CVF Winter Conference on Applications of Com- puter Vision (WACV), pages 1220–1230, 2022. 7, 8

    2022

  18. [26]

    Generative adversarial nets

    Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. InPro- ceedings of the Advances in Neural Information Processing Systems, 2014. 1, 2

    2014

  19. [27]

    Fisblim: A five-step blind metric for quality assessment of multiply distorted images

    Ke Gu, Guangtao Zhai, Min Liu, Xiaokang Yang, Wen- jun Zhang, Xianghui Sun, Wanhong Chen, and Ying Zuo. Fisblim: A five-step blind metric for quality assessment of multiply distorted images. InProceedings of the SiPS, pages 241–246, 2013. 7, 8

    2013

  20. [28]

    Vqfr: Blind face restoration with vector-quantized dictionary and parallel decoder.arXiv preprint arXiv:2205.06803, 2022

    Yuchao Gu, Xintao Wang, Liangbin Xie, Chao Dong, Gen Li, Ying Shan, and Ming-Ming Cheng. Vqfr: Blind face restoration with vector-quantized dictionary and parallel decoder.arXiv preprint arXiv:2205.06803, 2022. 2, 3, 1

    2022 arXiv

  21. [29]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 770–778, 2016. 7, 8

    2016

  22. [30]

    Id- animator: Zero-shot identity-preserving human video gen- eration.arXiv preprint arXiv:2404.15275, 2024

    Xuanhua He, Quande Liu, Shengju Qian, Xin Wang, Tao Hu, Ke Cao, Keyu Yan, Man Zhou, and Jie Zhang. Id- animator: Zero-shot identity-preserving human video gen- eration.arXiv preprint arXiv:2404.15275, 2024. 2

    2024 arXiv

  23. [31]

    Faceqnet: Quality assessment for face recognition based on deep learning

    Javier Hernandez-Ortega, Javier Galbally, Julian Fierrez, Rudolf Haraksim, and Laurent Beslay. Faceqnet: Quality assessment for face recognition based on deep learning. In 2019 International Conference on Biometrics (ICB), pages 1–8. IEEE, 2019. 2

    2019

  24. [32]

    Clipscore: A reference-free evaluation metric for image captioning.arXiv preprint arXiv:2104.08718, 2021

    Jack Hessel, Ari Holtzman, Maxwell Forbes, Ronan Le Bras, and Yejin Choi. Clipscore: A reference-free evaluation metric for image captioning.arXiv preprint arXiv:2104.08718, 2021. 7, 8

    2021 arXiv

  25. [33]

    Denoising dif- fusion probabilistic models.Proceedings of the Advances in Neural Information Processing Systems (NeurIPS), 33: 6840–6851, 2020

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models.Proceedings of the Advances in Neural Information Processing Systems (NeurIPS), 33: 6840–6851, 2020. 1

    2020

  26. [34]

    Image quality metrics: Psnr vs

    Alain Hore and Djemel Ziou. Image quality metrics: Psnr vs. ssim. InProceedings of the IEEE International Con- ference on Pattern Recoginition (ICPR), pages 2366–2369. IEEE, 2010. 7, 8

    2010

  27. [35]

    Lora: Low- rank adaptation of large language models

    Edward J Hu, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, Weizhu Chen, et al. Lora: Low- rank adaptation of large language models. InProceedings of the International Conference on Learning Representa- tions (ICLR), 2021. 2

    2021

  28. [36]

    Varfvv: View- adaptive real-time interactive free-view video streaming with edge computing.IEEE Journal on Selected Areas in Communications, 2025

    Qiang Hu, Qihan He, Houqiang Zhong, Guo Lu, Xiaoyun Zhang, Guangtao Zhai, and Yanfeng Wang. Varfvv: View- adaptive real-time interactive free-view video streaming with edge computing.IEEE Journal on Selected Areas in Communications, 2025. 2

    2025

  29. [37]

    4dgc: Rate-aware 4d gaussian compression for efficient streamable free-viewpoint video

    Qiang Hu, Zihan Zheng, Houqiang Zhong, Sihua Fu, Li Song, Xiaoyun Zhang, Guangtao Zhai, and Yanfeng Wang. 4dgc: Rate-aware 4d gaussian compression for efficient streamable free-viewpoint video. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition...

    2025

  30. [38]

    Huang, Manu Ramesh, Tamara Berg, and Erik Learned-Miller

    Gary B. Huang, Manu Ramesh, Tamara Berg, and Erik Learned-Miller. Labeled faces in the wild: A database for studying face recognition in unconstrained environ- ments. Technical Report 07-49, University of Mas- sachusetts, Amherst, 2007. 3

    2007

  31. [39]

    Kaiyi Huang, Kaiyue Sun, Enze Xie, Zhenguo Li, and Xi- hui Liu. T2i-compbench: A comprehensive benchmark for open-world compositional text-to-image generation.Pro- ceedings of the Advances in Neural Information Processing Systems (NeurIPS), 36:78723–78747, 2023. 2

    2023

  32. [40]

    Catekv: On sequential consistency for long-context llm inference acceleration

    Haoyun Jiang, Fei Huang, Qiang Hu, Minmin Sun, Shuai Xiao, Yong Li, Junyang Lin, Jiangchao Yao, et al. Catekv: On sequential consistency for long-context llm inference acceleration. InProceedings of International Conference on Machine Learning (ICML), 2025. 3

    2025

  33. [41]

    Ifqa: Interpretable face quality assessment

    Byungho Jo, Donghyeon Cho, In Kyu Park, and Sungeun Hong. Ifqa: Interpretable face quality assessment. InPro- ceedings of the IEEE/CVF Winter Conference on Applica- tions of Computer Vision (WACV), pages 3444–3453, 2023. 7, 8

    2023

  34. [42]

    Convo- lutional neural networks for no-reference image quality as- sessment

    Le Kang, Peng Ye, Yi Li, and David Doermann. Convo- lutional neural networks for no-reference image quality as- sessment. InProceedings of the IEEE Conference on Com- puter Vision and Pattern Recognition (CVPR), 2014. 7, 8

    2014

  35. [43]

    A style-based generator architecture for generative adversarial networks

    Tero Karras, Samuli Laine, and Timo Aila. A style-based generator architecture for generative adversarial networks. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR), pages 4401–4410,

  36. [44]

    Imagic: Text-based real image editing with diffusion mod- els

    Bahjat Kawar, Shiran Zada, Oran Lang, Omer Tov, Hui- wen Chang, Tali Dekel, Inbar Mosseri, and Michal Irani. Imagic: Text-based real image editing with diffusion mod- els. InProceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR), pages 6007–...

    2023

  37. [45]

    The megaface benchmark: 1 million faces for recognition at scale

    Ira Kemelmacher-Shlizerman, Steven M Seitz, Daniel Miller, and Evan Brossard. The megaface benchmark: 1 million faces for recognition at scale. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 4873–4882, 2016. 3

    2016

  38. [46]

    Progressive face super-resolution via attention to fa- cial landmark.arXiv preprint arXiv:1908.08239, 2019

    Deokyun Kim, Minseon Kim, Gihyun Kwon, and Dae-Shik Kim. Progressive face super-resolution via attention to fa- cial landmark.arXiv preprint arXiv:1908.08239, 2019. 2

    1908 arXiv

  39. [47]

    Auto-encoding variational bayes

    Diederik P Kingma, Max Welling, et al. Auto-encoding variational bayes. 2013. 1

    2013

  40. [48]

    Pick-a-pic: An open dataset of user preferences for text-to-image generation

    Yuval Kirstain, Adam Polyak, Uriel Singer, Shahbuland Matiana, Joe Penna, and Omer Levy. Pick-a-pic: An open dataset of user preferences for text-to-image generation. Proceedings of Advances in neural information processing systems (Neurips), 36:36652–36663, 2023. 1, 3

    2023

  41. [49]

    Multi-concept customization of text-to-image diffusion

    Nupur Kumari, Bingliang Zhang, Richard Zhang, Eli Shechtman, and Jun-Yan Zhu. Multi-concept customization of text-to-image diffusion. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 1931–1941, 2023. 2

    1931

  42. [51]

    Agiqa-3k: An open database for ai-generated image quality assessment.arXiv prepreint arXiv: 2306.04717,

    Chunyi Li, Zicheng Zhang, Haoning Wu, Wei Sun, Xiongkuo Min, Xiaohong Liu, Guangtao Zhai, and Weisi Lin. Agiqa-3k: An open database for ai-generated image quality assessment.arXiv prepreint arXiv: 2306.04717,

    arXiv

  43. [52]

    Playground v2.5: Three in- sights towards enhancing aesthetic quality in text-to-image generation, 2024

    Daiqing Li, Aleks Kamko, Ehsan Akhgari, Ali Sabet, Lin- miao Xu, and Suhail Doshi. Playground v2.5: Three in- sights towards enhancing aesthetic quality in text-to-image generation, 2024. 3, 1, 5

    2024

  44. [53]

    Blip: Bootstrapping language-image pre-training for uni- fied vision-language understanding and generation

    Junnan Li, Dongxu Li, Caiming Xiong, and Steven Hoi. Blip: Bootstrapping language-image pre-training for uni- fied vision-language understanding and generation. InPre- ceedings of International conference on machine learning (ICML), pages 12888–12900, 2022. 7, 8

    2022

  45. [54]

    Learning warped guidance for blind face restoration

    Xiaoming Li, Ming Liu, Yuting Ye, Wangmeng Zuo, Liang Lin, and Ruigang Yang. Learning warped guidance for blind face restoration. InThe European Conference on Computer Vision (ECCV), 2018. 2

    2018

  46. [55]

    Enhanced blind face restoration with multi-exemplar images and adaptive spatial feature fusion

    Xiaoming Li, Wenyu Li, Dongwei Ren, Hongzhi Zhang, Meng Wang, and Wangmeng Zuo. Enhanced blind face restoration with multi-exemplar images and adaptive spatial feature fusion. InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR), pages 2...

    2020

  47. [56]

    Learning dual memory dic- tionaries for blind face restoration.IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2022

    Xiaoming Li, Shiguang Zhang, Shangchen Zhou, Lei Zhang, and Wangmeng Zuo. Learning dual memory dic- tionaries for blind face restoration.IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2022. 3

    2022

  48. [57]

    Photomaker: Cus- tomizing realistic human photos via stacked id embedding

    Zhen Li, Mingdeng Cao, Xintao Wang, Zhongang Qi, Ming-Ming Cheng, and Ying Shan. Photomaker: Cus- tomizing realistic human photos via stacked id embedding. arXiv preprint arXiv:2312.04461, 2023. 1, 2, 3, 5

    2023 arXiv

  49. [58]

    Hunyuan- dit: A powerful multi-resolution diffusion transformer with fine-grained chinese understanding.arXiv preprint arXiv:2405.08748, 2024

    Zhimin Li, Jianwei Zhang, Qin Lin, Jiangfeng Xiong, Yanxin Long, Xinchi Deng, Yingfang Zhang, Xingchao Liu, Minbin Huang, Zedong Xiao, et al. Hunyuan- dit: A powerful multi-resolution diffusion transformer with fine-grained chinese understanding.arXiv preprint arXiv:2405.08748...

    2024 arXiv

  50. [59]

    Rich human feedback for text-to-image generation

    Youwei Liang, Junfeng He, Gang Li, Peizhao Li, Arseniy Klimovskiy, Nicholas Carolan, Jiao Sun, Jordi Pont-Tuset, Sarah Young, Feng Yang, et al. Rich human feedback for text-to-image generation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition...

    2024

  51. [60]

    Multi-branch face quality assess- ment for face recognition

    Zhang Lijun, Shao Xiaohu, Yang Fei, Deng Pingling, Zhou Xiangdong, and Shi Yu. Multi-branch face quality assess- ment for face recognition. In2019 IEEE 19th International Conference on Communication Technology (ICCT), pages 1659–1664. IEEE, 2019. 2

    2019

  52. [61]

    Microsoft coco: Common objects in context

    Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll ´ar, and C Lawrence Zitnick. Microsoft coco: Common objects in context. InProceedings of European conference on com- puter vision (ECCV), pages 740–755. Springer, 2014. 3

    2014

  53. [62]

    Diff- bir: Towards blind image restoration with generative diffu- sion prior.arXiv preprint arXiv:2308.15070, 2023

    Xinqi Lin, Jingwen He, Ziyan Chen, Zhaoyang Lyu, Ben Fei, Bo Dai, Wanli Ouyang, Yu Qiao, and Chao Dong. Diff- bir: Towards blind image restoration with generative diffu- sion prior.arXiv preprint arXiv:2308.15070, 2023. 1, 2, 3, 5

    2023 arXiv

  54. [63]

    Visual instruction tuning

    Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee. Visual instruction tuning. InProceedings of the Advances in Neural Information Processing Systems (NeurIPS), pages 34892–34916, 2023. 2

    2023

  55. [64]

    Blind quality assessment based on pseudo-reference image.IEEE Transactions on Multimedia (TMM), 20(8):2049–2062, 2018

    Xiongkuo Min, Ke Gu, Guangtao Zhai, Jing Liu, Xiaokang Yang, and Chang Wen Chen. Blind quality assessment based on pseudo-reference image.IEEE Transactions on Multimedia (TMM), 20(8):2049–2062, 2018. 7, 8

    2018

  56. [65]

    Blind image quality estimation via distortion aggravation.IEEE Transactions on Broadcasting, 64(2): 508–517, 2018

    Xiongkuo Min, Guangtao Zhai, Ke Gu, Yutao Liu, and Xi- aokang Yang. Blind image quality estimation via distortion aggravation.IEEE Transactions on Broadcasting, 64(2): 508–517, 2018. 8

    2018

  57. [66]

    Perceptual video quality assessment: A survey.Science China Information Sciences, 67(11): 211301, 2024

    Xiongkuo Min, Huiyu Duan, Wei Sun, Yucheng Zhu, and Guangtao Zhai. Perceptual video quality assessment: A survey.Science China Information Sciences, 67(11): 211301, 2024. 2

    2024

  58. [67]

    completely blind

    Anish Mittal, Rajiv Soundararajan, and Alan C. Bovik. Making a “completely blind” image quality analyzer.IEEE Signal Processing Letters, 20(3):209–212, 2013. 7, 8

    2013

  59. [68]

    Agedb: the first manually collected, in-the-wild age database

    Stylianos Moschoglou, Athanasios Papaioannou, Chris- tos Sagonas, Jiankang Deng, Irene Kotsia, and Stefanos Zafeiriou. Agedb: the first manually collected, in-the-wild age database. Inproceedings of the IEEE conference on computer vision and pattern recognition workshops, page...

    2017

  60. [69]

    Sdd-fiqa: Unsupervised face image quality assess- ment with similarity distribution distance

    Fu-Zhao Ou, Xingyu Chen, Ruixin Zhang, Yuge Huang, Shaoxin Li, Jilin Li, Yong Li, Liujuan Cao, and Yuan-Gen Wang. Sdd-fiqa: Unsupervised face image quality assess- ment with similarity distribution distance. InProceedings of the IEEE/CVF Conference on Computer Vision and Pat- ...

    2021

  61. [70]

    W ¨urstchen: An efficient architecture for large-scale text-to-image diffusion models

    Pablo Pernias, Dominic Rampas, Mats L Richter, Christo- pher J Pal, and Marc Aubreville. W ¨urstchen: An efficient architecture for large-scale text-to-image diffusion models. arXiv preprint arXiv:2306.00637, 2023. 3, 1, 4

    2023 arXiv

  62. [71]

    Sdxl: Improving latent diffusion mod- els for high-resolution image synthesis.arXiv preprint arXiv:2307.01952, 2023

    Dustin Podell, Zion English, Kyle Lacey, Andreas Blattmann, Tim Dockhorn, Jonas M ¨uller, Joe Penna, and Robin Rombach. Sdxl: Improving latent diffusion mod- els for high-resolution image synthesis.arXiv preprint arXiv:2307.01952, 2023. 2, 3, 5, 1, 4

    2023 arXiv

  63. [72]

    Zero-shot text-to-image generation, 2021

    Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea V oss, Alec Radford, Mark Chen, and Ilya Sutskever. Zero-shot text-to-image generation, 2021. 2

    2021

  64. [73]

    Hierarchical text-conditional image generation with clip latents.arXiv preprint arXiv:2204.06125, 2022

    Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen. Hierarchical text-conditional image generation with clip latents.arXiv preprint arXiv:2204.06125, 2022. 2

    2022 arXiv

  65. [74]

    High-resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Bj ¨orn Ommer. High-resolution image synthesis with latent diffusion models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 10684–10695, 2022. 1, 2, 3

    2022

  66. [75]

    Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation

    Nataniel Ruiz, Yuanzhen Li, Varun Jampani, Yael Pritch, Michael Rubinstein, and Kfir Aberman. Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages...

    2023

  67. [76]

    Sara Mahdavi, Rapha Gontijo Lopes, Tim Salimans, Jonathan Ho, David J Fleet, and Mohammad Norouzi

    Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily Denton, Seyed Kamyar Seyed Ghasemipour, Burcu Karagol Ayan, S. Sara Mahdavi, Rapha Gontijo Lopes, Tim Salimans, Jonathan Ho, David J Fleet, and Mohammad Norouzi. Photorealistic text-to- image diffusion mo...

  68. [77]

    Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily L Denton, Kamyar Ghasemipour, Raphael Gontijo Lopes, Burcu Karagol Ayan, Tim Sali- mans, et al. Photorealistic text-to-image diffusion mod- els with deep language understanding.Proceedings of the Advances...

    2022

  69. [78]

    Methodology for the subjective assessment of the quality of television pictures.Recommendation ITU-R BT, 500(13), 2012

    B Series. Methodology for the subjective assessment of the quality of television pictures.Recommendation ITU-R BT, 500(13), 2012. 4

    2012

  70. [79]

    Photo uncrop

    Qi Shan, Brian Curless, Yasutaka Furukawa, Carlos Her- nandez, and Steven M Seitz. Photo uncrop. InComputer Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part VI 13, pages 16–31. Springer, 2014. 7

    2014

  71. [80]

    Responsible re- search with crowds: pay crowdworkers at least minimum wage.Communications of the ACM, 61(3):39–41, 2018

    M Six Silberman, Bill Tomlinson, Rochelle LaPlante, Joel Ross, Lilly Irani, and Andrew Zaldivar. Responsible re- search with crowds: pay crowdworkers at least minimum wage.Communications of the ACM, 61(3):39–41, 2018. 2

    2018

  72. [81]

    Very Deep Con- volutional Networks for Large-Scale Image Recognition

    Karen Simonyan and Andrew Zisserman. Very Deep Con- volutional Networks for Large-Scale Image Recognition. arXiv e-prints, art. arXiv:1409.1556, 2014. 7, 8

    2014 arXiv

  73. [82]

    Denois- ing diffusion implicit models.arXiv:2010.02502, 2020

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denois- ing diffusion implicit models.arXiv:2010.02502, 2020. 1

    2010 arXiv

  74. [83]

    Blindly assess image qual- ity in the wild guided by a self-adaptive hyper network

    Shaolin Su, Qingsen Yan, Yu Zhu, Cheng Zhang, Xin Ge, Jinqiu Sun, and Yanning Zhang. Blindly assess image qual- ity in the wild guided by a self-adaptive hyper network. In 2020 IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR), pages 3664–3673, 2020. 7, 8

    2020

  75. [84]

    Going the extra mile in face image quality assess- ment: A novel database and model.IEEE Transactions on Multimedia, 2023

    Shaolin Su, Hanhe Lin, Vlad Hosu, Oliver Wiedemann, Jin- qiu Sun, Yu Zhu, Hantao Liu, Yanning Zhang, and Dietmar Saupe. Going the extra mile in face image quality assess- ment: A novel database and model.IEEE Transactions on Multimedia, 2023. 2, 3

    2023

  76. [85]

    Ser-fiq: Unsupervised estimation of face image quality based on stochastic embed- ding robustness

    Philipp Terhorst, Jan Niklas Kolf, Naser Damer, Florian Kirchbuchner, and Arjan Kuijper. Ser-fiq: Unsupervised estimation of face image quality based on stochastic embed- ding robustness. InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CV...

    2020

  77. [86]

    Towards all-in-one medical image re-identification

    Yuan Tian, Kaiyuan Ji, Rongzhao Zhang, Yankai Jiang, Chunyi Li, Xiaosong Wang, and Guangtao Zhai. Towards all-in-one medical image re-identification. InProceedings of the Computer Vision and Pattern Recognition Confer- ence, pages 30774–30786, 2025. 2

    2025

  78. [87]

    Stylegan2 distillation for feed-forward image manipulation

    Yuri Viazovetskyi, Vladimir Ivashkin, and Evgeny Kashin. Stylegan2 distillation for feed-forward image manipulation. Proceedings of the European conference on computer vi- sion (ECCV), pages 170–186, 2020. 2

    2020

  79. [88]

    Aigciqa2023: A large-scale im- age quality assessment database for ai generated images: from the perspectives of quality, authenticity and corre- spondence

    Jiarui Wang, Huiyu Duan, Jing Liu, Shi Chen, Xiongkuo Min, and Guangtao Zhai. Aigciqa2023: A large-scale im- age quality assessment database for ai generated images: from the perspectives of quality, authenticity and corre- spondence. InProceedings of the CAAI International Co...

    2023

  80. [89]

    Aigciqa2023: A large-scale im- age quality assessment database for ai generated images: from the perspectives of quality, authenticity and corre- spondence

    Jiarui Wang, Huiyu Duan, Jing Liu, Shi Chen, Xiongkuo Min, and Guangtao Zhai. Aigciqa2023: A large-scale im- age quality assessment database for ai generated images: from the perspectives of quality, authenticity and corre- spondence. 2023. 1, 2, 3, 9

    2023

  81. [90]

    Exploiting diffusion prior for real-world image super-resolution.arXiv preprint arXiv:2305.07015, 2023

    Jianyi Wang, Zongsheng Yue, Shangchen Zhou, Kelvin CK Chan, and Chen Change Loy. Exploiting diffusion prior for real-world image super-resolution.arXiv preprint arXiv:2305.07015, 2023. 1, 2, 3

    2023 arXiv

  82. [91]

    Understanding and evaluating human preferences for ai generated images with instruction tuning.arXiv preprint arXiv:2405.07346, 2024

    Jiarui Wang, Huiyu Duan, Guangtao Zhai, and Xiongkuo Min. Understanding and evaluating human preferences for ai generated images with instruction tuning.arXiv preprint arXiv:2405.07346, 2024. 7, 8

    2024 arXiv

  83. [92]

    Chan, and Chen Change Loy

    Jianyi Wang, Zongsheng Yue, Shangchen Zhou, Kelvin C.K. Chan, and Chen Change Loy. Exploit- ing diffusion prior for real-world image super-resolution

  84. [93]

    Xintao Wang, Honglun Zhang, Chao Dong, and Ying Shan

    Liangbin Xie. Xintao Wang, Honglun Zhang, Chao Dong, and Ying Shan. Vfhq: A high-quality dataset and bench- mark for video face super-resolution, 2022. 3

    2022

  85. [94]

    Instantid: Zero-shot identity-preserving gen- eration in seconds.arXiv preprint arXiv:2401.07519, 2024

    Qixun Wang, Xu Bai, Haofan Wang, Zekui Qin, and An- thony Chen. Instantid: Zero-shot identity-preserving gen- eration in seconds.arXiv preprint arXiv:2401.07519, 2024. 2, 3, 5, 1, 7

    2024 arXiv

  86. [95]

    A survey of deep face restoration: Denoise, super-resolution, deblur, artifact re- moval.arXiv preprint arXiv:2211.02831, 2022

    Tao Wang, Kaihao Zhang, Xuanxi Chen, Wenhan Luo, Jiankang Deng, Tong Lu, Xiaochun Cao, Wei Liu, Hong- dong Li, and Stefanos Zafeiriou. A survey of deep face restoration: Denoise, super-resolution, deblur, artifact re- moval.arXiv preprint arXiv:2211.02831, 2022. 2

    2022

  87. [96]

    Real-esrgan: Training real-world blind super-resolution with pure synthetic data

    Xintao Wang, Liangbin Xie, Chao Dong, and Ying Shan. Real-esrgan: Training real-world blind super-resolution with pure synthetic data. InInternational Conference on Computer Vision Workshops (ICCVW). 3, 1

  88. [97]

    To- wards real-world blind face restoration with generative fa- cial prior

    Xintao Wang, Yu Li, Honglun Zhang, and Ying Shan. To- wards real-world blind face restoration with generative fa- cial prior. InThe IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2021. 2, 3, 5, 1

    2021

  89. [98]

    To- wards real-world blind face restoration with generative fa- cial prior

    Xintao Wang, Yu Li, Honglun Zhang, and Ying Shan. To- wards real-world blind face restoration with generative fa- cial prior. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021. 3, 1

    2021

  90. [99]

    Image quality assessment: from error visibility to structural similarity.IEEE Transactions on Image Pro- cessing (TIP), 13(4):600–612, 2004

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Simoncelli. Image quality assessment: from error visibility to structural similarity.IEEE Transactions on Image Pro- cessing (TIP), 13(4):600–612, 2004. 7, 8

    2004

  91. [100]

    Dr2: Diffusion-based robust degradation remover for blind face restoration

    Zhixin Wang, Ziying Zhang, Xiaoyun Zhang, Huangjie Zheng, Mingyuan Zhou, Ya Zhang, and Yanfeng Wang. Dr2: Diffusion-based robust degradation remover for blind face restoration. InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR), pages ...

    2023

  92. [101]

    Elite: Encoding visual con- cepts into textual embeddings for customized text-to-image generation

    Yuxiang Wei, Yabo Zhang, Zhilong Ji, Jinfeng Bai, Lei Zhang, and Wangmeng Zuo. Elite: Encoding visual con- cepts into textual embeddings for customized text-to-image generation. InProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 15943– 1595...

    2023

  93. [102]

    Diffir: Efficient diffusion model for image restoration

    Bin Xia, Yulun Zhang, Shiyin Wang, Yitong Wang, Xing- long Wu, Yapeng Tian, Wenming Yang, and Luc Van Gool. Diffir: Efficient diffusion model for image restoration. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 13095–13105, 2023. 1

    2023

  94. [103]

    Tedigan: Text-guided diverse face image generation and manipulation

    Weihao Xia, Yujiu Yang, Jing-Hao Xue, and Baoyuan Wu. Tedigan: Text-guided diverse face image generation and manipulation. InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR), pages 2256–2265, 2021. 1

    2021

  95. [104]

    Fastcomposer: Tuning-free multi- subject image generation with localized attention.Inter- national Journal of Computer Vision (IJCV), 133(3):1175– 1194, 2025

    Guangxuan Xiao, Tianwei Yin, William T Freeman, Fr ´edo Durand, and Song Han. Fastcomposer: Tuning-free multi- subject image generation with localized attention.Inter- national Journal of Computer Vision (IJCV), 133(3):1175– 1194, 2025. 2, 3, 5, 1

    2025

  96. [105]

    Doermann

    Jingtao Xu, Peng Ye, Qiaohong Li, Haiqing Du, Yong Liu, and David S. Doermann. Blind image quality assessment based on high order statistics aggregation.IEEE Transac- tions on Image Processing (TIP), 25:4444–4457, 2016. 7, 8

    2016

  97. [106]

    Imagere- ward: Learning and evaluating human preferences for text- to-image generation.Proceedings of the Advances in Neu- ral Information Processing Systems (NeurIPS), 36, 2024

    Jiazheng Xu, Xiao Liu, Yuchen Wu, Yuxuan Tong, Qinkai Li, Ming Ding, Jie Tang, and Yuxiao Dong. Imagere- ward: Learning and evaluating human preferences for text- to-image generation.Proceedings of the Advances in Neu- ral Information Processing Systems (NeurIPS), 36, 2024. 2, 7, 8

    2024

  98. [107]

    Aigcoiqa2024: Perceptual qual- ity assessment of ai generated omnidirectional images

    Liu Yang, Huiyu Duan, Long Teng, Yucheng Zhu, Xiao- hong Liu, Menghan Hu, Xiongkuo Min, Guangtao Zhai, and Patrick Le Callet. Aigcoiqa2024: Perceptual qual- ity assessment of ai generated omnidirectional images. In 2024 IEEE International Conference on Image Processing (ICIP),...

    2024

  99. [108]

    Wider face: A face detection benchmark

    Shuo Yang, Ping Luo, Chen-Change Loy, and Xiaoou Tang. Wider face: A face detection benchmark. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 5525–5533, 2016. 3

    2016

  100. [109]

    MANIQA: Multi-dimension Attention Network for No- Reference Image Quality Assessment.arXiv e-prints, art

    Sidi Yang, Tianhe Wu, Shuwei Shi, Shanshan Lao, Yuan Gong, Mingdeng Cao, Jiahao Wang, and Yujiu Yang. MANIQA: Multi-dimension Attention Network for No- Reference Image Quality Assessment.arXiv e-prints, art. arXiv:2204.08958, 2022. 6, 7, 8

    2022 arXiv

  101. [110]

    Gan prior embedded network for blind face restoration in the wild

    Tao Yang, Peiran Ren, Xuansong Xie, and Lei Zhang. Gan prior embedded network for blind face restoration in the wild. InProceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition (CVPR), pages 672– 681, 2021. 2, 3, 1

    2021

  102. [111]

    Ip- adapter: Text compatible image prompt adapter for text-to- image diffusion models.arXiv preprint arXiv:2308.06721,

    Hu Ye, Jun Zhang, Sibo Liu, Xiao Han, and Wei Yang. Ip- adapter: Text compatible image prompt adapter for text-to- image diffusion models.arXiv preprint arXiv:2308.06721,

    arXiv

  103. [112]

    Difface: Blind face restoration with diffused error contraction.arXiv preprint arXiv:2212.06512, 2022

    Zongsheng Yue and Chen Change Loy. Difface: Blind face restoration with diffused error contraction.arXiv preprint arXiv:2212.06512, 2022. 2, 3, 1

    2022 arXiv

  104. [113]

    Perceptual image qual- ity assessment: a survey.Science China Information Sci- ences, 63:1–52, 2020

    Guangtao Zhai and Xiongkuo Min. Perceptual image qual- ity assessment: a survey.Science China Information Sci- ences, 63:1–52, 2020. 2

    2020

  105. [114]

    Lin Zhang, Lei Zhang, and Alan C. Bovik. A feature- enriched completely blind image quality evaluator.IEEE Transactions on Image Processing (TIP), 24(8):2579–2591,

  106. [115]

    The unreasonable effectiveness of deep features as a perceptual metric

    Richard Zhang, Phillip Isola, Alexei A Efros, Eli Shecht- man, and Oliver Wang. The unreasonable effectiveness of deep features as a perceptual metric. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 586–595, 2018. 6, 7, 8

    2018

  107. [116]

    anime” and “semi-realistic

    Shangchen Zhou, Kelvin Chan, Chongyi Li, and Chen Change Loy. Towards robust blind face restora- tion with codebook lookup transformer.Proceedings of the Advances in Neural Information Processing Systems (NeurIPS), 35:30599–30611, 2022. 2, 3, 5, 1, 7 F-Bench: Rethinking Human ...

    2022

  108. [117]

    QA Methods Implementation Details 11.1. Evaluation Metrics We adopt three widely used metrics in IQA [88, 89]: Spear- man rank-order correlation coefficient (SRCC), Pearson lin- ear correlation coefficient (PLCC), and Kendall rank corre- lation coefficient (KLCC) to evaluate t...

  109. [118]

    Authen.”, “Corres

    More Details of Our F-Eval Model 12.1. Loss Funtion We use both language loss, L1 loss and cross-entropy loss as the loss functions to optimize the training process. Specifi- cally, the language loss is used to restrict the F-Eval to pro- duce specific quality-related answer p...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.