Pith. sign in

REVIEW 4 major objections 5 minor 49 references

FakeIDet3-DB: Refining Digital Attacks and Patch Extraction for Secure ID Benchmarking

T0 review · 4 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read Current fake-ID detectors are far less reliable on real government-issued documents than on synthetic proxies, and this paper builds the first privacy-compliant benchmark to prove it.

desk verdict Valuable real-ID digital attack benchmark, but the privacy guarantee is unproven and the domain-gap claim rests on a confounded comparison. read the letter →

arxiv 2607.26641 v1 pith:RBXOA73Q submitted 2026-07-29 cs.CV cs.AI

Muñoz-Haro Javier , Teruel Andres , Tolosana Ruben , DeAlcala Daniel , Vera-Rodriguez Ruben , Morales Aythami , Fierrez Julian This is my paper
classification cs.CVcs.AI
keywords fakeidentitydocumentsdigitalforgerydetectiongenerativeAIattacksprivacy-preservingbenchmarkpatchextractionintegralimagenon-maximumsuppressiondomaingap
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

FakeIDet3-DB is the first database of digitally manipulated real, government-issued IDs, with over 6,400 images spanning classical and generative-AI attacks. The paper shows that state-of-the-art forgery detectors perform far worse on these real IDs than on synthetic templates: a leading detector's error roughly doubles. It also introduces PACE, a privacy-preserving algorithm that extracts patches from pseudo-anonymized IDs while concentrating on the residual information around redacted areas. The result is a public benchmark that lets researchers measure and improve detection without exposing personal data.

What carries the argument

The core mechanism is PACE (Pseudo-Anonymized Contextual patch Extraction), a three-stage algorithm that treats patch extraction as a constrained spatial packing problem. It uses an integral image (summed-area table) to verify in O(1) time whether a candidate patch touches any redacted pixel; a distance transform to rank valid patches by their proximity to anonymization-mask boundaries; and greedy spatial non-maximum suppression to place non-overlapping patches that 'crystallize' around redactions. This maximizes capture of residual sensitive data and manipulation artifacts while guaranteeing zero overlap with censored regions. The paper also relies on a refined attack-generation pipeline th

What would settle it

Run a re-identification attack on the released patches: attempt to match residual face fragments, text snippets, or ID numbers to public records. If any patch can be reliably linked to a specific person, the privacy guarantee fails and the database cannot be legally shared. Alternatively, evaluate the same detectors on a separate collection of real government IDs not touched by any preprocessing; if their error rates match synthetic-data performance, the domain-gap claim is refuted.

Watch

Extended reading notes

Core claim

On its own terms, the paper claims that current ID-forensics models are substantially less reliable on real government-issued documents than on synthetic proxies, and that this gap has been hidden by the lack of a legal, real-ID benchmark. The paper builds that benchmark and demonstrates the gap: one prominent detector's equal error rate jumps from about 16% on synthetic data to about 32% on real-ID attacks, while localization accuracy stays comparatively high. It argues that refined attacks—those post-processed with semantic alpha-blending and background inpainting—are much harder to detect than naive 'cheapfakes,' and that patch-based, privacy-compliant evaluation poses a major new challen

Load-bearing premise

The load-bearing premise is that pseudo-anonymization—leaving small residual sections of sensitive data uncovered—protects the ID owner's personally identifiable information; the paper provides no formal privacy analysis or empirical re-identification test.

Editorial extensions

If this is right

  • Detector rankings based on synthetic ID benchmarks may not transfer to real-world ID verification; real-ID evaluation is necessary.
  • The doubling of error rates on real IDs suggests current verification systems are more vulnerable to digital injection attacks than previously acknowledged.
  • Refined attacks degrade detector performance in over 80% of evaluated cases, so evaluation protocols should include refined forgeries rather than only cheapfakes.
  • Patch-based privacy-preserving benchmarks are viable, but existing detectors and naive score fusion are insufficient at small patch sizes; new architectures are needed.
  • The database enables quantitative study of domain shift between synthetic and real identity documents for forensics.

Reading between the lines

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

  • If the domain gap is as large as measured, regulatory and auditing standards for AI-based ID verification may need to require real-ID test sets before deployment.
  • PACE's redaction-boundary prioritization could generalize to other privacy-sensitive document types (passports, financial records, medical scans) where censored regions must remain inviolable.
  • The observation that localization outperforms detection on real-ID attacks hints that local forensic traces are present but global context masks them; training schemes that amplify local cues might close part of the gap.
  • A direct testable extension: fine-tuning detectors on FakeIDet3-DB patches should improve real-world ID forgery detection more than fine-tuning on synthetic data; the paper's own benchmark split enables this comparison.
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

4 major / 5 minor

Summary. FakeIDet3-DB introduces a new benchmark for digital manipulation detection on real, government-issued IDs, comprising 6,436 images from 250 real ID photos and 8 attack typologies, with a post-processing refinement pipeline for both classical and GenAI attacks. The paper also proposes PACE, a pseudo-anonymized contextual patch extraction algorithm based on integral images, distance transforms, and greedy NMS, and reports evaluations of several state-of-the-art detectors on full-image and patch-based benchmarks. The central claim is that FakeIDet3-DB is harder than existing synthetic proxies such as FantasyID, with TruFor's EER rising from 16.02% to 32.45%.

Significance. If the claims are supported, FakeIDet3-DB would be a valuable community resource: it addresses a real gap in the literature by providing publicly available, privacy-aware digital manipulations on genuine government IDs, and it ships benchmark code. PACE is a concrete algorithmic contribution to privacy-preserving patch extraction. The paper also provides reproducible evaluation protocols and makes the database available, which are strengths. However, the main evidence for the domain-gap claim is currently confounded, and the privacy guarantee is asserted rather than demonstrated. The resource is useful regardless, but the paper needs to either strengthen the causal evidence or substantially soften the headline claims.

major comments (4)
  1. [§V-A, Table V] The central claim that real government-issued IDs pose a harder detection challenge than synthetic proxies rests on Table V, where TruFor's EER rises from 16.02% on FantasyID to 32.45% on FakeIDet3-DB. This comparison is confounded: the two databases differ simultaneously in ID provenance, number and type of attack categories (8 vs 3), the presence of the refinement post-processing pipeline (Section III-B), image resolution, and sample size. The paper itself reports that refinement degrades detection in 83.33% of cases (§V-A), so the observed difference could be entirely attributable to attack sophistication rather than the real-ID substrate. A control experiment that applies the same attack pipeline and refinement to a synthetic template set, or restricts FakeIDet3-DB to the attack types present in FantasyID, is needed to isolate the effect of real ID structure. Without this, the headli
  2. [§IV, first paragraph] The privacy-compliance claim is load-bearing. The paper states that pseudo-anonymization leaves 'small, residual sections of the sensitive data uncovered which do not compromise the owners' PII,' but no formal leakage analysis, re-identification experiment, or quantitative privacy metric is provided. Since the entire distribution model and the legal validity of the benchmark depend on this assertion, the authors should provide evidence (e.g., information-theoretic guarantees, empirical re-identification risk on extracted patches, or compliance with a recognized anonymization standard) and specify what the residual sections contain. As written, the claim 'privacy-compliant benchmark' is unsubstantiated.
  3. [§IV-B, Table II] PACE is presented with 'optimal evaluation stride (s=2)' chosen from Table II, which is the same table used to claim superiority over baselines. This is a post-hoc model selection on the evaluation set. The difference between s=2 and s=1 is small (73.64 vs 73.16), and s=8 also outperforms baselines, so the point is not fatal, but the selection procedure should be described and ideally validated on a separate set or with a sensitivity analysis.
  4. [Abstract; §V-A, Tables III–V] The abstract's '32.45% EER in detection' is presented as a summary for 'they all struggle,' but this value is TruFor's detection EER on FakeIDet3-DB in Table V; Re-MTKD obtains 40.28% EER, while the fine-tuned Re-MTKD_ICCV25 attains much lower EER on GenAI attacks (12.18%, Table IV). Please qualify which model and which setting the headline numbers refer to. In addition, all evaluations are single-run with no confidence intervals or significance tests; with frozen weights this may be acceptable for many detectors, but the paper should state whether any stochasticity exists and report error bars where relevant.
minor comments (5)
  1. [Abstract] Typo: 'avaible' should be 'available'.
  2. [§IV-B] The text cites 'Kochkarev et al. [42]' where the reference number should be [13]; [42] is Chai et al.
  3. [§V-A] The statement 'we use the models' pre-trained weights frozen' appears only before Table V; it should be stated at the beginning of Section V for all evaluations.
  4. [Table V] The caption says 'comparison ... in terms of difficulty'; since the comparison is confounded, 'difficulty' should be described as database-level difficulty, not substrate-level difficulty.
  5. [Fig. 1] The figure legend is dense and difficult to parse; consider separating the attack-generation and patch-extraction pipelines into two figures or enlarging the labels.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity; benchmark evaluation is independent, with only minor self-citations and a same-table hyperparameter choice.

full rationale

The paper's central scientific claim—that FakeIDet3-DB is harder than FantasyID for frozen pre-trained detectors (TruFor EER 16.02% to 32.45%, Re-MTKD EER 33.67% to 40.28%, Table V)—is an empirical evaluation, not a derivation from fitted inputs. The detectors are evaluated with pre-trained, frozen weights against an external database (FantasyID), and no parameter of FakeIDet3-DB is fit to those detectors. PACE is presented as an optimization method; Table II measures how well it achieves the C_ROI and D_sem objectives that it was explicitly designed to maximize, and it is compared against external baselines (Kochkarev et al., PatchMatch). This is a self-evaluation against the algorithm's own objective, not a predictive claim, so it is not circular in the load-bearing sense. The choice of stride s=2 from the same Table II is a same-data hyperparameter selection; it is not presented as a prediction and does not force the benchmark outcomes. The paper builds on the authors' own prior privacy framework [11], [12] and reuses the 250 real IDs from FakeIDet2-db [12]; these self-citations are provenance and are not load-bearing for the detection-difficulty result. The weakest assumption is the assertion that residual pseudo-anonymized sections 'do not compromise the owners' PII' (Section IV), imported from [12] with no formal leakage analysis; this is a missing-support limitation, not a circular derivation. Likewise, the attribution of the difficulty gap to the real-ID substrate is confounded by attack refinement and differing attack typologies, but that is a validity/correctness concern, not a circularity. Overall, no step in the claimed derivation reduces to its own inputs by construction; the score of 2 reflects only minor self-citations and the same-table hyperparameter choice.

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

No new physical entities are introduced. The main load-bearing assumptions are the privacy claim and the applicability of the private source IDs. The free parameters are hand-chosen values in the attack refinement and patch extraction pipeline; the most consequential is the stride s=2 selected post hoc from the evaluation table.

free parameters (6)
  • Stride s = 2 (optimal)
    Chosen on the same 250-ID evaluation set in Table II; s=2 gives 73.64% ROI coverage vs 73.16% for s=1; declared optimal without error bars.
  • Patch sizes = 64x64, 128x128
    Two fixed sizes adopted from prior framework [12]; results differ substantially.
  • Gaussian kernel sizes for alpha mattes = 3x3 (text), 5x5 (face)
    Hand-chosen in Eqs. (3) and (5) to soften edges.
  • Morphological dilation kernel for face removal = 75x75
    Used in Section III-B.2 to erase the original face contour before background inpainting.
  • Blending factor λ in face morphing = unspecified (∈[0,1])
    Equation (1); controls morph strength, not set explicitly in paper.
  • Expanded center crop size for text inpainting = 768x768
    Chosen to provide receptive field for text generation (Section III-A.2).
assumptions (5)
  • domain assumption Pseudo-anonymization with residual sensitive data does not compromise PII
    Section IV: 'leaving small, residual sections ... which do not compromise the owners' PII' — no evidence.
  • domain assumption The 250 real IDs from FakeIDet2-db are government-issued and usable as ground truth
    Section III: 'we utilize the 250 real IDs introduced in FakeIDet2-db [12]'; no public verification.
  • domain assumption Ground-truth masks and bounding boxes are accurate
    Section IV-B: metrics depend on 'ground-truth ID bounding boxes' and manipulation masks; no annotation protocol or quality assessment.
  • domain assumption Frozen-weights SOTA models are representative baselines for detector capability
    Section V: evaluations use frozen weights; no fine-tuning protocol, so results may not reflect achievable performance.
  • standard math Standard image processing: integral image, distance transform, NMS
    Section IV-A; standard textbook tools.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FakeIDet3-DB: Refining Digital Attacks and Patch Extraction for Secure ID Benchmarking." pith.science (2026). https://pith.science/paper/RBXOA73Q

@misc{pith2026260726641,
  author       = {Pith},
  title        = {Pith review of: FakeIDet3-DB: Refining Digital Attacks and Patch Extraction for Secure ID Benchmarking},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RBXOA73Q}},
  note         = {Machine review of arXiv:2607.26641}
}
read the original abstract

Identity document (ID) authentication relies on the structural integrity of complex, high-frequency security patterns. However, advanced Generative AI models can now inject localized, high-fidelity manipulations, creating deceptive attacks that bypass standard verification. Training robust image forensic models to detect these anomalies is hindered by privacy regulations, forcing reliance on synthetic templates lacking the intricate visual patterns of real IDs. To bridge this domain gap, we introduce FakeIDet3-DB, the first comprehensive database of digital manipulations on real, government-issued IDs. FakeIDet3-DB encompasses classical (e.g., copy-move) and Generative AI-driven manipulations (e.g., face-swapping, inpainting) enhanced with advanced image refinement procedures to suppress visual artifacts. In addition, to comply with strict data protection regulations (e.g., GDPR), we adopt a recently-proposed framework based on patches. In order to maximize forensic utility, we formulate privacy-aware patch extraction from a real ID as a geometrically constrained image processing problem. We propose PACE, a Pseudo-Anonymized Contextual patch Extraction algorithm, which leverages Integral Image mapping and distance-driven Non-Maximum Suppression (NMS). PACE efficiently contours anonymization masks that prevent Personally Identifiable Information (PII) leakage while maximizing semantic density in peri-censorship regions, yielding almost 5.2M patches extracted from more than 6.4K images from real/fake IDs. Furthermore, an extensive evaluation of the proposed FakeIDet3-DB is performed using state-of-the-art models, showcasing they all struggle to detect and locate attacks coming from generative and classic techniques (32.45\% EER in detection and 83.48\% AUC-ROC in localization).

Figures

Figures reproduced from arXiv: 2607.26641 by the authors.

Figure 1
Figure 1. An overview of the main contributions of the present article. The left panel illustrates the attack generation pipeline considered in FakeIDet3-DB, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Visual results of the digital attacks included in FakeIDet3-DB including different types of manipulations and post-processing techniques. It is noticeable [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Proposed refined post-processing pipeline for text digital attacks. The GenAI pipeline (top, light-blue background) involves GenAI methods ( [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Proposed refined post-processing pipeline for face digital attacks. Given two images, source ( [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Visual example of a pseudo-anonymized ID, where redacted sections make practically impossible to infer the sensitive information from the owner, thus preventing PII leaking. Uncovered sections are denoted as residual sensitive data, which are important for targeted pat…
Figure 6
Figure 6. Figure 6: Visual examples of different patch extractions regarding ROI Coverage [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: Comparison in terms of valid extracted 128 [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

49 extracted references · 1 linked inside Pith

  1. [1]

    Leveraging Avatar Fingerprinting: A Multi-Generator Photorealistic Talking-Head Public Database and Benchmark,

    L. Pedrouzo-Rodriguezet al., “Leveraging Avatar Fingerprinting: A Multi-Generator Photorealistic Talking-Head Public Database and Benchmark,”Pattern Recognition, 2026

    2026

  2. [2]

    Deepfakes and Beyond: A Survey of Face Manip- ulation and Fake Detection,

    R. Tolosanaet al., “Deepfakes and Beyond: A Survey of Face Manip- ulation and Fake Detection,”Information Fusion, vol. 64, pp. 131–148, 2020

    2020

  3. [3]

    Face Morphing Attack Generation and Detection: A Comprehensive Survey,

    S. Venkateshet al., “Face Morphing Attack Generation and Detection: A Comprehensive Survey,”IEEE Technol. Soc. Mag., vol. 2, no. 3, pp. 128–145, 2021

    2021

  4. [4]

    Character-Aware Models Improve Visual Text Rendering,

    R. Liuet al., “Character-Aware Models Improve Visual Text Rendering,” inProc. 61st Annual Meeting of the Association for Computational Linguistics (ACL), 2023, pp. 16 270–16 297

    2023

  5. [5]

    Focus on Finding Deepfakes: A Robust Proactive Detection Method Based on Orthogonal Moment Watermarking,

    C. Wanget al., “Focus on Finding Deepfakes: A Robust Proactive Detection Method Based on Orthogonal Moment Watermarking,”IEEE Trans. Image Process., vol. 35, 2026

    2026

  6. [6]

    Toward Generalizable Forgery Detection and Reasoning,

    Y . Gaoet al., “Toward Generalizable Forgery Detection and Reasoning,” IEEE Trans. Image Process., vol. 35, pp. 3395–3410, 2026

    2026

  7. [7]

    Unsupervised Domain Adaptation-Based Cross-Type Deepfake Image Detection,

    Q. Wanget al., “Unsupervised Domain Adaptation-Based Cross-Type Deepfake Image Detection,”IEEE Trans. Image Process., vol. 35, pp. 4411–4424, 2026

    2026

  8. [8]

    FantasyID: A Dataset for Detecting Digital Manipulations in ID-Documents,

    P. Korshunovet al., “FantasyID: A Dataset for Detecting Digital Manipulations in ID-Documents,” inProc. IEEE/IAPR Intl. Joint Conf. on Biometrics (IJCB), 2025, pp. 1–9

    2025

Show all 49 references
  1. [9]

    DeepID Challenge of Detecting Synthetic Manipulations in ID Documents,

    ——, “DeepID Challenge of Detecting Synthetic Manipulations in ID Documents,” inProc. IEEE/CVF Intl. Conf. on Computer Vision (ICCV) Workshops, 2025, pp. 521–530

    2025

  2. [10]

    Second Competition on Presentation Attack Detection on ID Card,

    J. E. Tapiaet al., “Second Competition on Presentation Attack Detection on ID Card,” inProc. IEEE/IAPR Intl. Joint Conf. on Biometrics (IJCB), 2025, pp. 1–10

    2025

  3. [11]

    FakeIDet: Exploring Patches for Privacy- Preserving Fake ID Detection,

    J. Mu ˜noz-Haroet al., “FakeIDet: Exploring Patches for Privacy- Preserving Fake ID Detection,” inProc. IEEE/IAPR Intl. Joint Conf. on Biometrics (IJCB), 2025, pp. 1–9

    2025

  4. [12]

    Privacy-Aware Detection of Fake Identity Documents: Method- ology, Benchmark, and Improved Algorithms (FakeIDet2),

    ——, “Privacy-Aware Detection of Fake Identity Documents: Method- ology, Benchmark, and Improved Algorithms (FakeIDet2),”Information Fusion, vol. 128, p. 103969, 2026

    2026

  5. [13]

    Data Balancing Method for Training Segmentation Neural Networks,

    A. Kochkarevet al., “Data Balancing Method for Training Segmentation Neural Networks,” inProc. CEUR Workshop, 2020

    2020

  6. [14]

    KID34K: A Dataset for Online Identity Card Fraud Detection,

    E.-J. Parket al., “KID34K: A Dataset for Online Identity Card Fraud Detection,” inProc. ACM Intl. Conf. on Information and Knowledge Management, 2023, pp. 5381–5385

    2023

  7. [15]

    MIDV-2019: Challenges of the Modern Mobile-Based Document OCR,

    K. Bulatovet al., “MIDV-2019: Challenges of the Modern Mobile-Based Document OCR,” inProc. Intl. Conf. on Machine Vision (ICMV), 2020, pp. 818–824

    2019

  8. [16]

    MIDV-2020: A Comprehensive Benchmark Dataset for Identity Document Analysis,

    ——, “MIDV-2020: A Comprehensive Benchmark Dataset for Identity Document Analysis,”Computer Optics, vol. 46, 2022

    2020

  9. [17]

    MIDV-500: A Dataset for Identity Document Analysis and Recognition on Mobile Devices in Video Stream,

    V . V . Arlazarovet al., “MIDV-500: A Dataset for Identity Document Analysis and Recognition on Mobile Devices in Video Stream,”Com- puter Optics, vol. 43, 2019

    2019

  10. [18]

    Document Liveness Challenge Dataset (DLC- 2021),

    D. V . Polevoyet al., “Document Liveness Challenge Dataset (DLC- 2021),”Journal of Imaging, vol. 8, no. 7, 2022

    2021

  11. [19]

    IDNet: A Novel Identity Document Dataset via Few- Shot and Quality-Driven Synthetic Data Generation,

    L. Xieet al., “IDNet: A Novel Identity Document Dataset via Few- Shot and Quality-Driven Synthetic Data Generation,” inProc. IEEE Intl. Conf. on Big Data, 2024, pp. 2244–2253

    2024

  12. [20]

    Detecting Presentation Attacks on ID Cards Using Feature Refinement,

    R. Mudgalgunduraoet al., “Detecting Presentation Attacks on ID Cards Using Feature Refinement,” in2025 33rd European Signal Processing Conf. (EUSIPCO), 2025, pp. 825–829

    2025

  13. [21]

    Synthetic ID Card Image Generation for Improv- ing Presentation Attack Detection,

    D. Benalcazaret al., “Synthetic ID Card Image Generation for Improv- ing Presentation Attack Detection,”IEEE Trans. Inf. Forensics Security, vol. 18, pp. 1814–1824, 2023

    2023

  14. [22]

    PatchMatch: A Randomized Correspondence Algo- rithm for Structural Image Editing,

    C. Barneset al., “PatchMatch: A Randomized Correspondence Algo- rithm for Structural Image Editing,”ACM Trans. Graph., vol. 28, no. 3, p. 24, 2009

    2009

  15. [23]

    Scaling In-Context Segmentation with Hierarchical Supervision,

    T. C. Ndiret al., “Scaling In-Context Segmentation with Hierarchical Supervision,” inProc. IEEE/CVF Conf. on Computer Vision and Pattern Recognition Workshops (CVPRw), 2026, pp. 6150–6156

    2026

  16. [24]

    Image forgery detection,

    H. Farid, “Image forgery detection,”IEEE Signal Process. Mag., vol. 26, no. 2, pp. 16–25, 2009

    2009

  17. [25]

    Diffusion Models in Vision: A Survey,

    F.-A. Croitoruet al., “Diffusion Models in Vision: A Survey,”IEEE Trans. Pattern Anal. Mach. Intell., vol. 45, no. 9, pp. 10 850–10 869, 2023

    2023

  18. [26]

    Improving Diffusion Models for Scene Text Editing with Dual Encoders,

    J. Jiet al., “Improving Diffusion Models for Scene Text Editing with Dual Encoders,” arXiv:2304.05568, 2023

    2023 arXiv

  19. [27]

    UDiffText: A Unified Framework for High-Quality Text Synthesis in Arbitrary Images via Character-Aware Diffusion Models,

    Y . Zhaoet al., “UDiffText: A Unified Framework for High-Quality Text Synthesis in Arbitrary Images via Character-Aware Diffusion Models,” inProc. European Conf. on Computer Vision (ECCV), 2024, p. 217–233

    2024

  20. [28]

    TextDiffuser-2: Unleashing the Power of Language Models for Text Rendering,

    J. Chenet al., “TextDiffuser-2: Unleashing the Power of Language Models for Text Rendering,” inProc. European Conf. on Computer Vision (ECCV), 2025, pp. 386–402

    2025

  21. [29]

    Recognition-Synergistic Scene Text Editing,

    Z. Fanget al., “Recognition-Synergistic Scene Text Editing,” inProc. IEEE/CVF Conf. on Computer Vision and Pattern Recognition (CVPR), 2025, pp. 13 104–13 113

    2025

  22. [30]

    DiffMorpher: Unleashing the Capability of Diffusion Models for Image Morphing,

    K. Zhanget al., “DiffMorpher: Unleashing the Capability of Diffusion Models for Image Morphing,” inProc. IEEE/CVF Conf. on Computer Vision and Pattern Recognition (CVPR), 2024, pp. 7912–7921

    2024

  23. [31]

    FreeMorph: Tuning-Free Generalized Image Morphing with Diffusion Model,

    Y . Caoet al., “FreeMorph: Tuning-Free Generalized Image Morphing with Diffusion Model,” inProc. IEEE/CVF Intl. Conf. on Computer Vision (ICCV), 2025, pp. 18 111–18 120

    2025

  24. [32]

    Realistic and Efficient Face Swapping: A Unified Approach with Diffusion Models,

    S. Baliahet al., “Realistic and Efficient Face Swapping: A Unified Approach with Diffusion Models,” inProc. IEEE/CVF Winter Conf. on Applications of Computer Vision (WACV), 2025, pp. 1062–1071

    2025

  25. [33]

    A Review on Generative Adversarial Networks: Al- gorithms, Theory, and Applications,

    J. Guiet al., “A Review on Generative Adversarial Networks: Al- gorithms, Theory, and Applications,”IEEE Trans. Knowl. Data Eng., vol. 35, no. 4, pp. 3313–3332, 2023

    2023

  26. [34]

    FaceDancer: Pose- and Occlusion-Aware High Fi- delity Face Swapping,

    F. Rosberget al., “FaceDancer: Pose- and Occlusion-Aware High Fi- delity Face Swapping,” inProc. IEEE/CVF Winter Conf. on Applications of Computer Vision (WACV), January 2023, pp. 3454–3463

    2023

  27. [35]

    Resolution-robust Large Mask Inpainting with Fourier Convolutions,

    R. Suvorovet al., “Resolution-robust Large Mask Inpainting with Fourier Convolutions,” inProc. IEEE/CVF Winter Conf. on Applications of Computer Vision (WACV), 2022, pp. 3172–3182

    2022

  28. [36]

    CNN-Generated Images Are Surprisingly Easy to Spot... for Now,

    S.-Y . Wanget al., “CNN-Generated Images Are Surprisingly Easy to Spot... for Now,” inProc. IEEE/CVF Conf. on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 8695–8704

    2020

  29. [37]

    Deep Image Fingerprint: Towards Low Budget Synthetic Image Detection and Model Lineage Analysis,

    S. Sinitsa and O. Fried, “Deep Image Fingerprint: Towards Low Budget Synthetic Image Detection and Model Lineage Analysis,” inProc. IEEE/CVF Winter Conf. on Applications of Computer Vision (WACV), 2024, pp. 4067–4076

    2024

  30. [38]

    GANprintR: Improved Fakes and Evaluation of the State of the Art in Face Manipulation Detection,

    J. C. Neveset al., “GANprintR: Improved Fakes and Evaluation of the State of the Art in Face Manipulation Detection,”IEEE J. Sel. Topics Signal Process., pp. 1038–1048, 2020

    2020

  31. [39]

    Hi-SAM: Marrying Segment Anything Model for Hier- archical Text Segmentation,

    M. Yeet al., “Hi-SAM: Marrying Segment Anything Model for Hier- archical Text Segmentation,”IEEE Trans. Pattern Anal. Mach. Intell., vol. 47, no. 03, pp. 1431–1447, 2025

    2025

  32. [40]

    SAM 2: Segment anything in images and videos,

    N. Raviet al., “SAM 2: Segment anything in images and videos,” in Proc. Intl. Conf. on Learning Representations (ICLR), 2025

    2025

  33. [41]

    Learning Patch-Channel Correspondence for Interpretable Face Forgery Detection,

    Y . Huaet al., “Learning Patch-Channel Correspondence for Interpretable Face Forgery Detection,”IEEE Trans. Image Process., vol. 32, pp. 1668– 1680, 2023

    2023

  34. [42]

    What Makes Fake Images Detectable? Understanding Properties that Generalize,

    L. Chaiet al., “What Makes Fake Images Detectable? Understanding Properties that Generalize,” inProc. European Conf. on Computer Vision (ECCV), 2020

    2020

  35. [43]

    Sequential Operations in Digital Picture Processing,

    A. Rosenfeld and J. L. Pfaltz, “Sequential Operations in Digital Picture Processing,”J. ACM, vol. 13, no. 4, p. 471–494, 1966

    1966

  36. [44]

    Distance Functions on Digital Pictures,

    A. Rosenfeld and J. Pfaltz, “Distance Functions on Digital Pictures,” Pattern Recognition, vol. 1, no. 1, pp. 33–61, 1968

    1968

  37. [45]

    First Competition on Presentation Attack Detection on ID Card,

    J. E. Tapiaet al., “First Competition on Presentation Attack Detection on ID Card,” inProc. IEEE/IAPR Intl. Joint Conf. on Biometrics (IJCB), 2024, pp. 1–10

    2024

  38. [46]

    TruFor: Leveraging All-Round Clues for Trustworthy Image Forgery Detection and Localization,

    F. Guillaroet al., “TruFor: Leveraging All-Round Clues for Trustworthy Image Forgery Detection and Localization,” inProc. IEEE/CVF Conf. on Computer Vision and Pattern Recognition (CVPR), June 2023, pp. 20 606–20 615

    2023

  39. [47]

    DiffForensics: Leveraging Diffusion Prior to Image Forgery Detection and Localization,

    Z. Yuet al., “DiffForensics: Leveraging Diffusion Prior to Image Forgery Detection and Localization,” inProc. IEEE/CVF Conf. on Computer Vision and Pattern Recognition (CVPR), 2024, pp. 12 765–12 774

    2024

  40. [48]

    Reinforced Multi-teacher Knowledge Distillation for Efficient General Image Forgery Detection and Localization,

    ——, “Reinforced Multi-teacher Knowledge Distillation for Efficient General Image Forgery Detection and Localization,” inProc. AAAI Conf. on Artificial Intelligence, vol. 39, 2025, pp. 995–1003

    2025

  41. [49]

    Can We Get Rid of Handcrafted Feature Extractors? Spar- seViT: Nonsemantics-Centered, Parameter-Efficient Image Manipulation Localization Through Spare-Coding Transformer,

    L. Suet al., “Can We Get Rid of Handcrafted Feature Extractors? Spar- seViT: Nonsemantics-Centered, Parameter-Efficient Image Manipulation Localization Through Spare-Coding Transformer,” inProc. AAAI Conf. on Artificial Intelligence, vol. 39, no. 7, 2025, pp. 7024–7032

    2025

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.