Pith. sign in

REVIEW 3 major objections 5 minor 78 references

Medical Manifestation-Aware De-Identification

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

Pith's one-line read Synthetic patient faces could let doctors share images without leaking identity or losing disease clues.

desk verdict First public synthetic medical face dataset with a reversible DeID method that survives an external clinical trial; the privacy guarantee is weaker than the abstract claims. read the letter →

arxiv 2412.10804 v1 pith:Z5T5AE24 submitted 2024-12-14 cs.CV cs.AI

classification cs.CVcs.AI
keywords facede-identificationmedicalmanifestationpreservationsyntheticpatientfacesdiffusionmodelsreversibleimageprivacyeyediseasedataset
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 argues that face de-identification for medical images fails because existing methods erase the disease signs doctors need, and because no large patient-face dataset exists for learning those signs. It answers both problems at once: it releases MeMa, a 42,307-image synthetic dataset re-generated from real patient photos and annotated by clinicians, and it proposes MedSem-DeID, a reversible de-identification pipeline that injects a medical-semantics prior learned from MeMa into the encryption process. The reported outcome is that de-identified faces keep their diagnostic value (86.70% disease classification, 0.6775 segmentation Dice, physician agreement kappa above 0.81) while identity matching falls to 1.76% in a simulated authentication system. If correct, this would make privacy-protected sharing of patient face imagery practical for medical AI.

What carries the argument

The load-bearing mechanism is MedSem-DeID, a pipeline whose image-encoder feature map is channel-concatenated with a medical feature map extracted by the frozen early blocks of a diffusion model trained on MeMa, then encrypted by password-conditioned Transformer blocks (ID-Encryptor) and decrypted by matching Transformer blocks (ID-Decryptor). The medical prior enters twice: as a feature-level injection and as a loss term Lmed that forces the encrypted image to match the original's medical semantics. A second diffusion model fine-tuned with LoRA on real patient data, sampled with IP-Adapter injections from public faces and real-guided age/gender/disease distributions, produces MeMa itself.

What would settle it

Run a re-identification study using newer photographs of the 12,467 original patients against all 42,307 published MeMa images, or apply a stronger face-recognition model than the three used; if any real patient can be reliably re-identified from a published MeMa image, the central privacy claim fails.

Watch

Extended reading notes

Core claim

The central claim is that medical manifestations are learnable, preservable, and separable from identity: a diffusion model fine-tuned on real patient photos and re-sampled with public-face attribute injection yields a large synthetic patient cohort whose disease signs physicians rate as plausible, and a de-identification network guided by a frozen medical-semantics encoder can hide identity while keeping those exact signs. The paper further claims that this dataset-plus-prior recipe outperforms all previous de-identification methods on both medical utility and privacy metrics, including on a real-world cohort of 129 patients from a different hospital.

Load-bearing premise

The claim that MeMa never leaks real patients' identities rests entirely on three face-recognition models failing to match the generated images back to the original photos.

Editorial extensions

If this is right

  • Disease classification on de-identified faces reaches 86.70% accuracy, more than ten points above the best prior method.
  • Tumor segmentation on de-identified images reaches a Dice score of 0.6775, beating both hand-crafted medical priors and generic utility-preserving methods.
  • Real physicians agree with original diagnoses at kappa above 0.81, even for disease types not seen during training.
  • De-identified faces match the real patient's ID photo only 1.76% of the time in a simulated identity-authentication system.
  • The method is reversible: with the correct password the original image is recovered with 89.34% disease-classification accuracy and 27.02 dB PSNR.

Reading between the lines

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

  • If the privacy claim survives a stronger re-identification attack, the MeMa recipe could be ported to other sensitive image domains, such as dermatology or psychiatry, without new patient photography.
  • The 1.27% residual identity-match rate is the natural stress point: a re-identification study using newer photographs of the same 12,467 patients would reveal whether those matches are true look-alikes or memorized identities that the three matchers happened to miss.
  • Because MedSem-DeID is reversible by design, its practical privacy depends on password custody: anyone holding the password can reconstruct the original image, so the method fits controlled clinical workflows where traceability is legally required rather than public release of encrypted images.
  • The medical-semantics encoder trained on MeMa could transfer to other ocular-image tasks, such as lesion synthesis or disease-progression simulation, since it already encodes localized diagnostic cues as dense features.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 addresses face de-identification in medical scenes. The authors construct MeMa, a dataset of 42,307 synthetic patient-face images generated from 39,323 real patient photos via a fine-tuned Stable Diffusion model with IP-Adapter injection of public face attributes, followed by identity-distance filtering and physician quality filtering and annotation. They propose MedSem-DeID, a reversible DeID method that injects a medical semantics encoder (trained on MeMa) into a GAN-based encryption/decryption pipeline and regularizes the de-identified image to preserve medical features. Experiments on MeMa show state-of-the-art disease classification (86.70% Top-1) and segmentation Dice (0.6775), and a real-world clinical trial on 129 patients reports physician diagnostic consistency kappa > 0.81.

Significance. If the results hold, the paper makes a valuable contribution by introducing the first large-scale synthetic patient-face dataset (MeMa) with expert annotations and a reversible DeID method that explicitly preserves medical manifestation semantics. The inclusion of a real-world clinical trial on an external cohort (ECXHCSU) is a notable strength, as is the planned release of dataset and code, which will facilitate reproducibility. However, the main benchmark evaluation is subject to a closed-loop between the medical-utility classifier and the proposed method's medical prior, and the privacy guarantee for MeMa is supported only by empirical matching rates without formal analysis. These issues need to be resolved before the quantitative claims can be fully accepted.

major comments (3)
  1. [Experiments / Evaluation Protocol and Metrics] The main medical-utility evaluation is circular. The DiNov2 classifier is fine-tuned on the MeMa training set (Section 'Evaluation Protocol and Metrics'), and the proposed MedSem-DeID method obtains its medical prior from the MeMa-trained medical encoder (Section 'Medical Semantics Encoding'). Both the classifier and the encoder are therefore adapted to the same synthetic image distribution, so the 86.70% Top-1 accuracy on the MeMa validation set may reflect distribution alignment rather than preservation of clinically meaningful manifestations. To support the claim that MedSem-DeID preserves medical utility, the paper should evaluate on an independent real-image benchmark, for instance a classifier trained on real patient photographs, or report the real-world physician diagnosis as the primary utility metric rather than only as a secondary outcome.
  2. [Approach / MeMa Dataset (Steps 2-3, Table 1)] The privacy claim that 'MeMa avoids breaching real patient privacy' (Abstract) is not established by the presented evidence. The reduction in identity-leakage percentage from 71.8% to 1.27% is measured with three face-recognition models (Table 1), but no formal privacy analysis (e.g., differential privacy, membership inference) is provided. More importantly, the residual 1.27% is not analyzed: the paper does not report how many images were removed by the Step-3 identity-distance filter, whether any residual matching images remain in the released 42,307-image set, or what re-identification risk those residual images pose. Additionally, the face-recognition models used are trained on ordinary face photographs and may not detect memorization of medical-specific image cues. The authors should either provide a rigorous privacy evaluation or substantially temper the strong privacy guarantee in the abstract and dataset description.
  3. [MeMa Dataset (Step 3) vs. Experiments (Datasets)] There is a factual inconsistency in the description of the MeMa-Seg subset. Step 3 of the dataset construction says the physicians segment the tumor mask on the subset of 'SCC images', while the Experiments section defines MeMa-Seg as 'for the BCC (basal cell carcinoma) disease type' (and the table captions refer to MeMa-Seg without specifying the disease). This discrepancy must be corrected, as it is essential for reproducibility and for interpreting the segmentation results (Table 3, Dice 0.6775).
minor comments (5)
  1. [Table 8 caption] The table header reads 'MaMa validation set'; this should be 'MeMa validation set'.
  2. [Figure 8] The axis labels in Figure 8 appear garbled in the supplied version; please ensure the figure is rendered correctly in the final version.
  3. [References] The citation to the MSD manual is incomplete; a full bibliographic entry should be provided.
  4. [Table 5] The model names 'FaceNet VGGFace2' and 'FaceNetCASIA' should have consistent formatting (e.g., spaces or underscores) for readability.
  5. [Conclusion and Limitation] The limitation section only mentions the narrow scope of disease types; it should also acknowledge the limitations of the privacy analysis and the potential for distribution shift between synthetic and real patient data.

Circularity Check

1 steps flagged · score 4.0 of 10

Partial circularity in the MeMa medical-utility benchmark; independent real-world trial prevents full reduction.

  1. fitted input called prediction [Approach: Medical Semantics Encoding; Learning Objectives; Experiments: Evaluation Protocol and Metrics]
    "we train another diffusion model on the proposed MeMa dataset to learn the medical semantics. ... Lmed = ℓ2(fmed, Encmed(Xenc)) facilitate the encrypted image is similar to the original image in terms of medical semantics. ... for the disease classification task, we fine-tune the DiNov2 model on the MeMa training set. We evaluate its Top1 accuracy on the MeMa validation set processed by various DeID approaches."

    Both the medical prior Encmed and the medical-utility probe DiNov2 are fitted to the same MeMa training distribution, and the MedSem-DeID objective explicitly trains Encmed(Xenc) to match fmed(X). The headline Med-Class numbers in Table 3 therefore measure how well de-identified images stay inside the MeMa-learned semantic space shared by the encoder and the classifier, rather than providing an independent check that diagnosis-relevant signs survive. This is a closed evaluation loop: the method is optimized to preserve the features of a MeMa-trained encoder, and utility is then scored by a MeMa-trained classifier.

full rationale

The paper's central derivation is a dataset-construction and benchmark pipeline, not a formal first-principles prediction. The main circularity concern is that the medical utility evaluation is partly self-referential: the medical semantics encoder and the DiNov2 classifier are both trained on MeMa, and the method is trained with Lmed to match the encoder's features, so high Med-Class accuracy on the MeMa validation set partly reflects alignment of training distributions rather than independently verified medical preservation. This loop is materially mitigated by the real-world clinical trial on Real-ECXHCSU with physician Cohen's kappa >0.81, which is outside the MeMa training distribution and still shows strong performance. The self-citations in the paper are background motivation (e.g., diffusion semantics references) and are not load-bearing; no uniqueness theorem or ansatz is imported from the authors' prior work, and no result is renamed as a new discovery. The identity-privacy claim rests on an empirical face-matching metric and filtering procedure, which is a definitional safeguard rather than a circular derivation, though the 1.27% residual and lack of formal privacy bounds remain a correctness risk rather than a circularity risk. Overall, the benchmark loop is a real but partial circularity, and the independent external trial keeps the central claim from collapsing into it.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The central claims rest on two domain assumptions: synthetic faces can stand in for real patients, and face-recognition match rate is an adequate privacy proxy. The method also has two hand-tuned loss weights selected on a validation split. No new physical entities are introduced.

free parameters (3)
  • lambda_med = 5
    Chosen on the 3,729-image hyper-parameter selection set to trade off medical classification accuracy and ID-Dis (Figure 8, left). The central claim of medical utility preservation depends on this balance.
  • lambda_rev = 0.1
    Chosen on the hyper-parameter selection set to balance reversible reconstruction quality and DeID performance (Figure 8, right). The reversibility claim depends on this weight.
  • IP-Adapter injection weight range = [0.2, 0.4] random
    Hand-set range to improve diversity and reduce identity leakage; not fitted to data but a design choice affecting the privacy and diversity trade-off.
assumptions (3)
  • domain assumption Synthetic faces generated by a LoRA-fine-tuned Stable Diffusion with IP-Adapter injection are sufficiently representative of real patient faces to train and evaluate medical DeID systems.
    The entire MeMa dataset and all models trained on it rely on this transfer; the paper provides indirect evidence via physician filtering and the Real-ECXHCSU trial, but it is not proven.
  • domain assumption Identity leakage can be measured by the match rate against SphereFace, ArcFace, and CosFace; no formal privacy guarantee is provided.
    The claim that MeMa avoids breaching patient privacy rests on this empirical proxy (Table 1), not on a formal re-identification risk analysis.
  • domain assumption The DiNov2 classifier fine-tuned on MeMa is a valid proxy for medical utility in the benchmark.
    Med-Class accuracy on MeMa validation is used as the primary medical utility metric; its validity depends on the synthetic training distribution matching real clinical relevance.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Medical Manifestation-Aware De-Identification." pith.science (2026). https://pith.science/paper/Z5T5AE24

@misc{pith2026241210804,
  author       = {Pith},
  title        = {Pith review of: Medical Manifestation-Aware De-Identification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Z5T5AE24}},
  note         = {Machine review of arXiv:2412.10804}
}
read the original abstract

Face de-identification (DeID) has been widely studied for common scenes, but remains under-researched for medical scenes, mostly due to the lack of large-scale patient face datasets. In this paper, we release MeMa, consisting of over 40,000 photo-realistic patient faces. MeMa is re-generated from massive real patient photos. By carefully modulating the generation and data-filtering procedures, MeMa avoids breaching real patient privacy, while ensuring rich and plausible medical manifestations. We recruit expert clinicians to annotate MeMa with both coarse- and fine-grained labels, building the first medical-scene DeID benchmark. Additionally, we propose a baseline approach for this new medical-aware DeID task, by integrating data-driven medical semantic priors into the DeID procedure. Despite its conciseness and simplicity, our approach substantially outperforms previous ones. Dataset is available at https://github.com/tianyuan168326/MeMa-Pytorch.

Figures

Figures reproduced from arXiv: 2412.10804 by the authors.

Figure 1
Figure 1. (a) Common DeID approaches, focus on remov [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Examples and the distribution characteristics of the proposed MeMa dataset. [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. MeMa building pipeline. (a) Training patient face [PITH_FULL_IMAGE:figures/full_fig_p002_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Comparison of different image generation strate [PITH_FULL_IMAGE:figures/full_fig_p003_4.png]
Figure 5
Figure 5. Figure 5: Overview of the proposed baseline model MedSem-DeID. The snow icon indicates the Encmed is frozen during training DeID networks. The image decoder after the ID-decryptor is omitted for briefness. ⊕ denotes the channel-wise concatenation operation. Given the original im…
Figure 6
Figure 6. Figure 6: Qualitative results of different methods on the MeMA validation set. [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]
Figure 7
Figure 7. Figure 7: Qualitative results of the recovered image of dif [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8: Impact of the loss weights for the medical ( [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

78 extracted references · 48 canonical work pages

  1. [1]

    A.; Hempel, T.; Khalifa, A.; Al-Hamadi, A.; and Dinges, L

    Abdelrahman, A. A.; Hempel, T.; Khalifa, A.; Al-Hamadi, A.; and Dinges, L. 2023. L2cs-net: Fine-grained gaze estimation in unconstrained environments. In 2023 8th International Conference on Frontiers of Signal Processing (ICFSP), 98--102. IEEE

  2. [2]

    Banerjee, M.; Capozzoli, M.; McSweeney, L.; and Sinha, D. 1999. Beyond kappa: A review of interrater agreement measures. Canadian journal of statistics, 27(1): 3--23

  3. [3]

    S.; and Wu, Z

    Cai, Z.; Gao, Z.; Planche, B.; Zheng, M.; Chen, T.; Asif, M. S.; and Wu, Z. 2024. Disguise without disruption: Utility-preserving face de-identification. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, 918--926

  4. [4]

    Cao, J.; Liu, B.; Wen, Y.; Xie, R.; and Song, L. 2021. Personalized and invertible face de-identification by disentangled identity information manipulation. In Proceedings of the IEEE/CVF international conference on computer vision, 3334--3342

  5. [5]

    M.; and Zisserman, A

    Cao, Q.; Shen, L.; Xie, W.; Parkhi, O. M.; and Zisserman, A. 2018. Vggface2: A dataset for recognising faces across pose and age. In 2018 13th IEEE international conference on automatic face & gesture recognition (FG 2018), 67--74. IEEE

  6. [6]

    Chen, H.; Qu, Z.; Tian, Y.; Jiang, N.; Qin, Y.; Gao, J.; Zhang, R.; Ma, Y.; Jin, Z.; and Zhai, G. 2024 a . A cross-temporal multimodal fusion system based on deep learning for orthodontic monitoring. Computers in Biology and Medicine, 180: 109025

  7. [7]

    Chen, Z.; Sun, W.; Tian, Y.; Jia, J.; Zhang, Z.; Wang, J.; Huang, R.; Min, X.; Zhai, G.; and Zhang, W. 2024 b . GAIA: Rethinking Action Quality Assessment for AI-Generated Videos. arXiv preprint arXiv:2406.06087

  8. [8]

    Deng, J.; Guo, J.; Xue, N.; and Zafeiriou, S. 2019. Arcface: Additive angular margin loss for deep face recognition. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 4690--4699

Show all 78 references
  1. [9]

    Duan, H.; Shen, W.; Min, X.; Tian, Y.; Jung, J.-H.; Yang, X.; and Zhai, G. 2022. Develop then rival: A human vision-inspired framework for superimposed image decomposition. IEEE Transactions on Multimedia, 25: 4267--4281

  2. [10]

    Z.; et al

    Fletcher, S.; Islam, M. Z.; et al. 2018. Comparing sets of patterns with the Jaccard index. Australasian Journal of Information Systems, 22

  3. [11]

    Gao, C.; Jiang, Y.; Wu, S.; Ma, Y.; Li, L.; and Liu, D. 2024. IMOFC: Identity-Level Metric Optimized Feature Compression for Identification Tasks. IEEE Transactions on Circuits and Systems for Video Technology

  4. [12]

    Gao, C.; Li, L.; Liu, D.; Chen, Z.; Li, W.; and Wu, F. 2022. Two-step fast mode decision for intra coding of screen content. IEEE Transactions on Circuits and Systems for Video Technology, 32(8): 5608--5622

  5. [13]

    Gao, C.; Liu, D.; Li, L.; and Wu, F. 2021. Towards task-generic image compression: A study of semantics-oriented metrics. IEEE Transactions on Multimedia, 25: 721--735

  6. [14]

    Gross, R.; Sweeney, L.; De la Torre, F.; and Baker, S. 2006. Model-based face de-identification. In 2006 Conference on computer vision and pattern recognition workshop (CVPRW'06), 161--161. IEEE

  7. [15]

    S.; and Lee, Y

    Gu, X.; Luo, W.; Ryoo, M. S.; and Lee, Y. J. 2020. Password-conditioned anonymization and deanonymization with face identity transformers. In European conference on computer vision, 727--743. Springer

  8. [16]

    Guideline, I. H. T. 2001. Guideline for good clinical practice. J Postgrad Med, 47(3): 199--203

  9. [17]

    He, K.; Chen, X.; Xie, S.; Li, Y.; Doll \'a r, P.; and Girshick, R. 2022. Masked autoencoders are scalable vision learners. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 16000--16009

  10. [18]

    He, K.; Fan, H.; Wu, Y.; Xie, S.; and Girshick, R. 2020. Momentum contrast for unsupervised visual representation learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 9729--9738

  11. [19]

    He, K.; Zhang, X.; Ren, S.; and Sun, J. 2016. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, 770--778

  12. [20]

    Hedlin, E.; Sharma, G.; Mahajan, S.; Isack, H.; Kar, A.; Tagliasacchi, A.; and Yi, K. M. 2024. Unsupervised semantic correspondence using stable diffusion. Advances in Neural Information Processing Systems, 36

  13. [21]

    J.; Shen, Y.; Wallis, P.; Allen-Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W

    Hu, E. J.; Shen, Y.; Wallis, P.; Allen-Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W. 2021. Lora: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685

  14. [22]

    Hu, Z.; Li, S.; Zhang, C.; Yi, K.; Wang, G.; and Manocha, D. 2020. Dgaze: Cnn-based gaze prediction in dynamic scenes. IEEE transactions on visualization and computer graphics, 26(5): 1902--1911

  15. [23]

    B.; Mattar, M.; Berg, T.; and Learned-Miller, E

    Huang, G. B.; Mattar, M.; Berg, T.; and Learned-Miller, E. 2008 a . Labeled faces in the wild: A database forstudying face recognition in unconstrained environments. In Workshop on faces in'Real-Life'Images: detection, alignment, and recognition

  16. [24]

    B.; Mattar, M.; Berg, T.; and Learned-Miller, E

    Huang, G. B.; Mattar, M.; Berg, T.; and Learned-Miller, E. 2008 b . Labeled Faces in the Wild: A Database forStudying Face Recognition in Unconstrained Environments. Month

  17. [25]

    Hukkel s, H.; Mester, R.; and Lindseth, F. 2019. Deepprivacy: A generative adversarial network for face anonymization. In International symposium on visual computing, 565--578. Springer

  18. [26]

    F.; Kohl, S

    Isensee, F.; Jaeger, P. F.; Kohl, S. A.; Petersen, J.; and Maier-Hein, K. H. 2021. nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nature methods, 18(2): 203--211

  19. [27]

    Jourabloo, A.; Yin, X.; and Liu, X. 2015. Attribute preserved face de-identification. In 2015 International conference on biometrics (ICB), 278--285. IEEE

  20. [28]

    F.; Simpson, J

    Kamnitsas, K.; Ledig, C.; Newcombe, V. F.; Simpson, J. P.; Kane, A. D.; Menon, D. K.; Rueckert, D.; and Glocker, B. 2017. Efficient multi-scale 3D CNN with fully connected CRF for accurate brain lesion segmentation. Medical image analysis, 36: 61--78

  21. [29]

    Karkkainen, K.; and Joo, J. 2021. Fairface: Face attribute dataset for balanced race, gender, and age for bias measurement and mitigation. In Proceedings of the IEEE/CVF winter conference on applications of computer vision, 1548--1558

  22. [30]

    Karras, T.; Aila, T.; Laine, S.; and Lehtinen, J. 2017. Progressive growing of gans for improved quality, stability, and variation. arXiv preprint arXiv:1710.10196

  23. [31]

    Karras, T.; Laine, S.; and Aila, T. 2019. A style-based generator architecture for generative adversarial networks. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 4401--4410

  24. [32]

    King, D. E. 2009. Dlib-ml: A machine learning toolkit. The Journal of Machine Learning Research, 10: 1755--1758

  25. [33]

    Kingma, D. 2014. Adam: a method for stochastic optimization. arXiv preprint arXiv:1412.6980

  26. [34]

    C.; Lo, W.-Y.; et al

    Kirillov, A.; Mintun, E.; Ravi, N.; Mao, H.; Rolland, C.; Gustafson, L.; Xiao, T.; Whitehead, S.; Berg, A. C.; Lo, W.-Y.; et al. 2023. Segment anything. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 4015--4026

  27. [35]

    Kong, Y.; and Fu, Y. 2022. Human action recognition and prediction: A survey. International Journal of Computer Vision, 130(5): 1366--1401

  28. [36]

    Li, C.; Zhang, J.; Zhang, Z.; Wu, H.; Tian, Y.; Sun, W.; Lu, G.; Liu, X.; Min, X.; Lin, W.; et al. 2024 a . R-Bench: Are your Large Multimodal Model Robust to Real-world Corruptions? arXiv preprint arXiv:2410.05474

  29. [37]

    Li, D.; Wang, W.; Zhao, K.; Dong, J.; and Tan, T. 2023. RiDDLE: Reversible and Diversified De-Identification With Latent Encryptor. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 8093--8102

  30. [38]

    Li, X.; Lu, J.; Han, K.; and Prisacariu, V. A. 2024 b . Sd4match: Learning to prompt stable diffusion model for semantic matching. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 27558--27568

  31. [39]

    Liu, W.; Wen, Y.; Yu, Z.; Li, M.; Raj, B.; and Song, L. 2017. Sphereface: Deep hypersphere embedding for face recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, 212--220

  32. [40]

    Maximov, M.; Elezi, I.; and Leal-Taix \'e , L. 2020. Ciagan: Conditional identity anonymization generative adversarial networks. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 5447--5456

  33. [41]

    Merck & Co., R. N. U., Inc. 2024. MSD MANUALS: The Trusted Provider of Medical Information since 1899. https://www.msdmanuals.com/

  34. [42]

    Min, X.; Duan, H.; Sun, W.; Zhu, Y.; and Zhai, G. 2024. Perceptual video quality assessment: A survey. Science China Information Sciences, 67(11): 211301

  35. [43]

    Mohsenin, A.; and Huang, J. J. 2012. Ocular manifestations of systemic inflammatory diseases. Connecticut medicine, 76(9)

  36. [44]

    Oquab, M.; Darcet, T.; Moutakanni, T.; Vo, H.; Szafraniec, M.; Khalidov, V.; Fernandez, P.; Haziza, D.; Massa, F.; El-Nouby, A.; et al. 2023. Dinov2: Learning robust visual features without supervision. arXiv preprint arXiv:2304.07193

  37. [45]

    N.; and Cohen, I

    Price, W. N.; and Cohen, I. G. 2019. Privacy in the age of medical big data. Nature medicine, 25(1): 37--43

  38. [46]

    J.; and Ryoo, M

    Ren, Z.; Lee, Y. J.; and Ryoo, M. S. 2018. Learning to anonymize faces for privacy preserving action detection. In Proceedings of the european conference on computer vision (ECCV), 620--636

  39. [47]

    Rombach, R.; Blattmann, A.; Lorenz, D.; Esser, P.; and Ommer, B. 2022. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 10684--10695

  40. [48]

    R \"u schendorf, L. 1985. The Wasserstein distance and approximation theorems. Probability Theory and Related Fields, 70(1): 117--129

  41. [49]

    Schroff, F.; Kalenichenko, D.; and Philbin, J. 2015. Facenet: A unified embedding for face recognition and clustering. In Proceedings of the IEEE conference on computer vision and pattern recognition, 815--823

  42. [50]

    Schuhmann, C.; Beaumont, R.; Vencu, R.; Gordon, C.; Wightman, R.; Cherti, M.; Coombes, T.; Katta, A.; Mullis, C.; Wortsman, M.; et al. 2022. Laion-5b: An open large-scale dataset for training next generation image-text models. Advances in Neural Information Processing Systems,...

  43. [51]

    Serengil, S. 2020. DeepFace: Face Recognition with Deep Neural Networks. https://github.com/serengil/deepface

  44. [52]

    I.; and Ozpinar, A

    Serengil, S. I.; and Ozpinar, A. 2021. HyperExtended LightFace: A Facial Attribute Analysis Framework. In 2021 International Conference on Engineering and Emerging Technologies (ICEET), 1--4. IEEE

  45. [53]

    Sharir, G.; Noy, A.; and Zelnik-Manor, L. 2021. An image is worth 16x16 words, what is a video worth? arXiv preprint arXiv:2103.13915

  46. [54]

    Tan, X.; Zhu, Y.; Cheng, Z.; Hu, M.; Zhang, X.; Pei, G.; Yu, C.; Li, Q.; Li, W.; and Wang, J. 2024. Low-cost and portable physiological signal monitor using PhysRate model. Displays, 81: 102605

  47. [55]

    P.; and Hariharan, B

    Tang, L.; Jia, M.; Wang, Q.; Phoo, C. P.; and Hariharan, B. 2023. Emergent correspondence from image diffusion. Advances in Neural Information Processing Systems, 36: 1363--1389

  48. [56]

    Tian, J.; Aggarwal, L.; Colaco, A.; Kira, Z.; and Gonzalez-Franco, M. 2024 a . Diffuse Attend and Segment: Unsupervised Zero-Shot Segmentation using Stable Diffusion. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 3554--3563

  49. [57]

    Tian, Y.; Che, Z.; Bao, W.; Zhai, G.; and Gao, Z. 2020. Self-supervised motion representation via scattering local motion cues. In Computer Vision--ECCV 2020: 16th European Conference, Glasgow, UK, August 23--28, 2020, Proceedings, Part XIV 16, 71--89. Springer

  50. [58]

    Tian, Y.; Lu, G.; Min, X.; Che, Z.; Zhai, G.; Guo, G.; and Gao, Z. 2021. Self-conditioned probabilistic learning of video rescaling. In Proceedings of the IEEE/CVF international conference on computer vision, 4490--4499

  51. [59]

    Tian, Y.; Lu, G.; Yan, Y.; Zhai, G.; Chen, L.; and Gao, Z. 2024 b . A coding framework and benchmark towards low-bitrate video understanding. IEEE Transactions on Pattern Analysis and Machine Intelligence

  52. [60]

    Tian, Y.; Lu, G.; and Zhai, G. 2024. SMC++: Masked Learning of Unsupervised Video Semantic Compression. arXiv preprint arXiv:2406.04765

  53. [61]

    Tian, Y.; Lu, G.; and Zhai, G. 2025. Free-VSC: Free Semantics from Visual Foundation Models for Unsupervised Video Semantic Compression. In European Conference on Computer Vision, 163--183. Springer

  54. [62]

    Tian, Y.; Lu, G.; Zhai, G.; and Gao, Z. 2023 a . Non-semantics suppressed mask learning for unsupervised video semantic compression. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 13610--13622

  55. [63]

    Tian, Y.; Min, X.; Zhai, G.; and Gao, Z. 2019. Video-based early asd detection via temporal pyramid networks. In 2019 IEEE International Conference on Multimedia and Expo (ICME), 272--277. IEEE

  56. [64]

    Tian, Y.; Yan, Y.; Zhai, G.; Chen, L.; and Gao, Z. 2023 b . Clsa: a contrastive learning framework with selective aggregation for video rescaling. IEEE Transactions on Image Processing, 32: 1300--1314

  57. [65]

    Tian, Y.; Yan, Y.; Zhai, G.; Guo, G.; and Gao, Z. 2022. Ean: event adaptive network for enhanced action recognition. International Journal of Computer Vision, 130(10): 2453--2471

  58. [66]

    N.; Kaiser, .; and Polosukhin, I

    Vaswani, A.; Shazeer, N.; Parmar, N.; Uszkoreit, J.; Jones, L.; Gomez, A. N.; Kaiser, .; and Polosukhin, I. 2017. Attention is all you need. Advances in neural information processing systems, 30

  59. [67]

    Wang, H.; Wang, Y.; Zhou, Z.; Ji, X.; Gong, D.; Zhou, J.; Li, Z.; and Liu, W. 2018. Cosface: Large margin cosine loss for deep face recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, 5265--5274

  60. [68]

    Wen, Y.; Liu, B.; Cao, J.; Xie, R.; and Song, L. 2023. Divide and conquer: a two-step method for high quality face de-identification with model explainability. In Proceedings of the IEEE/CVF International Conference on Computer Vision, 5148--5157

  61. [69]

    Yan, Z.; Li, S.; Zhao, R.; Tian, Y.; and Zhao, Y. 2023. DHBE: data-free holistic backdoor erasing in deep neural networks via restricted adversarial distillation. In Proceedings of the 2023 ACM Asia Conference on Computer and Communications Security, 731--745

  62. [70]

    Yang, Y.; Lyu, J.; Wang, R.; Wen, Q.; Zhao, L.; Chen, W.; Bi, S.; Meng, J.; Mao, K.; Xiao, Y.; et al. 2022. A digital mask to safeguard patient privacy. Nature medicine, 28(9): 1883--1892

  63. [71]

    Ye, H.; Zhang, J.; Liu, S.; Han, X.; and Yang, W. 2023. Ip-adapter: Text compatible image prompt adapter for text-to-image diffusion models. arXiv preprint arXiv:2308.06721

  64. [72]

    Yi, D.; Lei, Z.; Liao, S.; and Li, S. Z. 2014. Learning face representation from scratch. arXiv preprint arXiv:1411.7923

  65. [73]

    Yi, F.; Chen, M.; Sun, W.; Min, X.; Tian, Y.; and Zhai, G. 2021. Attention based network for no-reference UGC video quality assessment. In 2021 IEEE international conference on image processing (ICIP), 1414--1418. IEEE

  66. [74]

    Yi, X.; Jiang, Q.; and Zhou, W. 2024. No-reference quality assessment of underwater image enhancement. Displays, 81: 102586

  67. [75]

    Zhang, J.; Huang, J.; Jin, S.; and Lu, S. 2024. Vision-language models for vision tasks: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence

  68. [76]

    A.; Shechtman, E.; and Wang, O

    Zhang, R.; Isola, P.; Efros, A. A.; Shechtman, E.; and Wang, O. 2018. The unreasonable effectiveness of deep features as a perceptual metric. In Proceedings of the IEEE conference on computer vision and pattern recognition, 586--595

  69. [77]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all...

  70. [78]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

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