Pith. sign in

REVIEW 4 major objections 4 minor 27 references

A three-stage synthetic-image pipeline trains a skin-cancer classifier to 90.9% accuracy on real clinical images, beating prior state-of-the-art while narrowing skin-tone fairness gaps. Starting from only 656 real biopsy-confirmed images, t

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-02 06:08 UTC pith:HDHQIDJD

load-bearing objection The framework and the open 266k-image release are real contributions, but the headline accuracy numbers rest on a leakage-prevention sentence that is too thin to trust, and the test sets are too small for the claimed precision. the 4 major comments →

arxiv 2607.12987 v2 pith:HDHQIDJD submitted 2026-07-14 cs.CV

Controllable Generation of Diverse Dermatological Imagery for Fair and Efficient Malignancy Classification

H\'ector Carri\'on , Narges Norouzi This is my paper
classification cs.CV
keywords dermatologyfairnesssynthetic datadiffusion modelslesion mappingtextual inversionmalignancy classificationskin-tone diversity
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper claims that a hybrid generative pipeline can largely replace real annotated dermatology data. From 656 real images, it produces 266k+ synthetic images across three complementary routes: healthy-skin inpainting, non-parametric lesion transplant, and parametric diffusion generation. Training a classifier on synthetics alone reaches 86.4% accuracy on the DDI benchmark; adding a real-data fine-tuning pass raises accuracy to 90.9%, above the previous best, while improving Equality of Opportunity from 69.6 to 86.6. The result matters because dark skin tones and rare diseases are exactly where annotated data is scarcest, and the method is designed to generate those cases.

Core claim

On the biopsy-confirmed DDI benchmark, synthetic-only training reaches 86.4% accuracy with substantially higher fairness metrics than prior methods; adding a fine-tuning pass on real images lifts accuracy to 90.9%, above the previous best, and raises Equality of Opportunity from 69.6 to 86.6. Accuracy holds even for diseases with only one or two real samples (83.3% from synthetic data alone), demonstrating that single-sample lesions can be turned into trainable data via non-parametric lesion mapping. Cross-dataset experiments show a +13.9 percentage-point accuracy improvement on unseen expert-verified data despite minimal disease overlap.

What carries the argument

The central mechanism is the three-stage cgDDI generator: a latent-diffusion inpainting step that removes lesions to produce healthy skin canvases; a non-parametric lesion-mapping algorithm that pastes a real donor lesion onto a healthy canvas at a controlled location and skin tone; and a parametric semantic generator that fine-tunes a latent diffusion model per disease using textual inversion and low-rank adaptation, anchored by a prior-preservation loss on the healthy synthetics to prevent semantic drift. The non-parametric route is what makes single-sample diseases augmentable, while the parametric route scales diversity for diseases with roughly ten or more samples.

Load-bearing premise

The load-bearing premise is that a synthetic image built from a donor lesion pasted on another patient's healthy skin retains the donor lesion's disease identity and malignancy label; if label preservation fails, especially for rare diseases or cross-skin-tone transplants, the synthetic training set contains label noise that could explain the reported accuracy and fairness gains.

What would settle it

Take a balanced sample of lesion-mapped synthetic images for each disease and skin tone, have board-certified dermatologists classify them, and measure agreement with the donor's original label; if agreement drops materially for cross-tone or rare-disease transplants, the label-preservation premise fails. A simpler quantitative check: train a classifier on synthetics where lesion labels are randomly permuted among skin tones and see if accuracy drops; if it does not, the classifier is using background or tone cues rather than lesion identity.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • If the label-preservation assumption holds, synthetic-only training can match or beat real-data baselines, relaxing the need for large expert-annotated datasets.
  • Single-sample rare diseases become trainable: the 1–2 sample disease class reaches 83.3% accuracy from synthetic data alone.
  • Fairness improves alongside accuracy: Equality of Opportunity rises from 69.6 to 86.6, indicating more consistent true-positive rates across skin tones.
  • Cross-dataset transfer improves accuracy by 13.9 percentage points on unseen data with minimal disease overlap, suggesting the synthetic images carry generalizable lesion features.
  • The released 266k+ synthetic images, code, and generative models let others reproduce or extend the training data without sharing patient images.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The key untested assumption is that a lesion transplanted onto a healthy canvas of a different skin tone and body site preserves its disease identity; a dermatologist labeling study on lesion-mapped samples would directly test this and could reveal hidden label noise.
  • The +13.9% cross-dataset gain hints that synthetic data may act as a domain-adaptation mechanism, but the paper does not isolate whether the improvement comes from lesion morphology or from simple skin-tone rebalancing.
  • The 'about 10 samples' requirement for parametric generation is a soft bound; a systematic sample-size sweep with stronger regularization could reveal whether 3–5 samples suffice, further lowering data needs.
  • Fairness gains in discriminative classifiers may transfer to other medical imaging domains with similar scarcity imbalances, but that transfer is not demonstrated here and would need its own validation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper introduces cgDDI, a hybrid generative framework that expands a 656-image dermatology dataset (DDI) into 266k+ synthetic images through three mechanisms: inpainting-based healthy skin synthesis, non-parametric lesion mapping for rare diseases, and parametric semantic generation via textual inversion plus LoRA with prior-preservation loss. The authors evaluate malignancy classification and skin-tone fairness on DDI and Fitzpatrick17k, reporting 86.4% synthetic-only and 90.9% fine-tuned accuracy on DDI, with Equality of Opportunity improving from 69.6 to 86.6, and cross-dataset gains on F17k. All code, models, and synthetic images are released.

Significance. If the results hold, this is a valuable contribution to synthetic-data-augmented medical imaging: it is a rare openly released large synthetic dermatology corpus, it demonstrates single-sample disease augmentation, and it applies prior-preservation loss with healthy in-distribution anchors, a sensible and transferable idea. The authors also provide fairness metrics beyond accuracy and test on two datasets with different masking protocols. The central claims are plausible but currently rest on several unverified assumptions about leakage prevention and label preservation, and the statistical evidence for the headline deltas is thin.

major comments (4)
  1. [Sec. 4.2 / Sec. 3.2–3.4] The leakage-prevention statement is not sufficient to rule out test-set leakage. The healthy synthetics are generated once from 334 masked sDDI inputs (Sec. 3.2), and all lesion-mapped and semantic synthetics are built on those healthy canvases (Secs. 3.3–3.4). If any healthy canvas originated from a DDI image that appears in a fold's test set, every downstream synthetic inherits background cues (ruler, skin texture, framing) from that test image. The paper does not describe per-fold generation, does not report how many synthetics were excluded by the stated rule, and does not state whether the 309 healthy synthetics were derived from training-fold images only. This is load-bearing: a ViT can exploit such background identity cues, inflating both Exp. 1 and Exp. 2. Please specify the exact exclusion procedure, report excluded counts, or regenerate per fold.
  2. [Sec. 4.1 / Table 2] The DDI test set is only 131 images per fold and the F17k hold-out is 55 images. The headline improvements (e.g., 87.4% to 90.9%, EOM 69.6 to 86.6) correspond to a handful of images, and no confidence intervals or significance tests are reported. Please report bootstrap CIs or per-fold paired tests across the five seeds, especially for the fairness metrics whose standard deviations in Table 2 are large (DPM ±118, EOM ±1.9–4.3). Without this, the SOTA claim is not statistically grounded.
  3. [Sec. 3.3] The non-parametric lesion mapping assumes that a lesion transplanted onto a healthy canvas preserves the donor's disease identity and malignancy label even when moved to a different skin tone and body location. This is not validated. The synthetic-only accuracy for 1–2 sample diseases (83.3%, Table 3) is based on only 6 test cases and could reflect label noise from failed transplantations. Please provide a human-expert or at least a per-disease, per-mapping-parameter evaluation of label preservation, and report per-disease synthetic-only accuracies with intervals.
  4. [Sec. 3.4 / Eq. (1)] There is an internal inconsistency about the number of diseases used for parametric generation. Sec. 3.4 says 'For diseases with ≥10 samples' but then defines D = {d_j}_{40}, which equals 13 diseases with >10 samples plus 27 diseases with 2–10 samples (Sec. 3.1). Since the total of 185,400 semantic synthetics uses D=40, diseases with as few as 2 samples were in fact used. This contradicts the abstract's 'as few as 10 training samples' phrasing and the later claim that ~10 samples are sufficient. Please clarify the threshold and, if 2–10 sample diseases were included, adjust the claims and report their generation quality separately.
minor comments (4)
  1. [Table 4] The formatting of Table 4 is hard to read: values are not aligned, no ±std are given, and several DPM/EOM entries are exactly 0.500, which should be explained (ties or degenerate predictions?).
  2. [Sec. 4.5 / Table 6] The generative quality metrics lack a real-data baseline or comparison; FID values around 88–108 are high in absolute terms. Reporting only max/min ratios across skin tones does not establish 'fairness' of generation quality.
  3. [Sec. 3.1 / Sec. 3.4] The relationship between the 65 consolidated disease labels, the 25 single-observation diseases, and the 40 diseases in Eq. (1) should be made explicit in one place; currently the reader must infer it.
  4. [Sec. 4.2] The phrase 'excluding training synthetics conditioned on downstream test images' should define 'conditioned on' precisely: does it exclude healthy canvases only, or also lesion donors and semantic prompts originating from test images?

Circularity Check

0 steps flagged

No circularity found; the reported classification gains are benchmark results against held-out real data with a stated leakage exclusion, and no load-bearing reduction to fitted inputs or self-citations is present.

full rationale

The paper's central claims are empirical benchmark results: synthetics-only training reaches 86.4% DDI accuracy and fine-tuning reaches 90.9%, evaluated on held-out real DDI images under a stated leakage-prevention rule (Sec. 4.2). The generation pipeline (inpainting to healthy synthetics, non-parametric lesion mapping, and textual-inversion/LoRA semantic generation) operates on inputs drawn from the training portion of the data: DDI masks, donor lesions, healthy canvases, and prompts. No equation defines a predicted quantity in terms of the measured quantity it is meant to predict; Eq. (1) is a generative sampling procedure, not an estimator whose output is forced by the evaluation labels. The one self-citation, sDDI [6], supplies segmentation masks for DDI images; this is a reusable dataset artifact rather than a load-bearing theorem or fitted parameter, and the paper additionally demonstrates algorithmic masking with SAMv3 on F17k, so the central derivation does not reduce to that citation. Potential test-set leakage through synthetic images conditioned on held-out images is a soundness/selection-bias concern, but the manuscript explicitly states that training synthetics conditioned on downstream test images were excluded (Sec. 4.2), and no internal evidence contradicts that statement. Accordingly, there is no definitional, parametric, or citation-forced circularity in the claimed derivation chain.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 0 invented entities

The paper introduces no physical or mathematical entities; its free parameters are experimental design choices (sample counts, hyperparameters), and its axioms are data-quality and label-preservation assumptions.

free parameters (4)
  • Semantic samples per triple R = 5
    Eq. (1) uses R=5 samples per (healthy image, disease, skin tone) triple; hand-chosen, directly sets the 185,400-image count.
  • Minimum samples for parametric generation = 10
    Sec. 3.4: 'We find ∼10 samples sufficient'; this threshold divides diseases into parametric vs. non-parametric handling.
  • Skin-tone bins = 3 (light/medium/dark)
    Sec. 4.1: Fitzpatrick I–VI collapsed to three bins for generation conditioning and all fairness metrics; affects PQD/DPM/EOM values.
  • Generation hyperparameters α, β, t, LoRA rank, PPL weight = not reported
    Sec. 3.4: conditioning strength, guidance scale, inference steps, LoRA rank, and prior-preservation weight are hand-chosen and not listed or ablated.
axioms (5)
  • domain assumption DDI biopsy-confirmed diagnoses and Fitzpatrick labels are accurate
    Used as ground truth for classification and fairness; Sec. 3.1 and Sec. 4.1.
  • domain assumption sDDI segmentation masks correctly delineate lesions
    Masks drive inpainting and lesion mapping for DDI; citation [6] is the authors' own prior work; Sec. 3.2.
  • domain assumption A lesion transplanted onto a different patient's healthy skin retains its disease identity and malignancy label
    Non-parametric mapping in Sec. 3.3 assumes no label distortion from changes in skin tone, location, or background; undermines synthetic training if false.
  • domain assumption Prior-preservation loss on healthy synthetics prevents semantic drift without introducing artifacts
    Sec. 3.4 relies on PPL for faithful semantic generation; no quantitative verification that drift is controlled beyond downstream accuracy.
  • domain assumption SAMv3 automated masks on F17k are accurate enough for generation
    Sec. 4.3 pipelines F17k through SAMv3; authors note masks can be noisy and are discarded more often, but the retained subset is assumed usable.

pith-pipeline@v1.3.0-alltime-deepseek · 8150 in / 12123 out tokens · 103694 ms · 2026-08-02T06:08:14.101572+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of Controllable Generation of Diverse Dermatological Imagery for Fair and Efficient Malignancy Classification." pith.science (2026). https://pith.science/paper/HDHQIDJD

@misc{pith2026260712987,
  author       = {Pith},
  title        = {Pith review of: Controllable Generation of Diverse Dermatological Imagery for Fair and Efficient Malignancy Classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HDHQIDJD}},
  note         = {Machine review of arXiv:2607.12987}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Accurate dermatological diagnosis naturally necessitates equitable performance across diverse populations, yet a systematic lack of expertly annotated images, especially for underrepresented skin tones and rare diseases, impedes progress toward measurably fair methods. We introduce cgDDI (Controllable Generation of Diverse Dermatological Imagery), a hybrid framework that (1) synthesizes realistic healthy skin samples without disturbing other input properties, (2) maps single-sample rare lesions onto novel skin-tones and locations non-parametrically, and (3) allows for efficient parametric generation with as few as 10 training samples. The framework supports both human and automated segmentation masking, enabling scalability to datasets without pre-made lesion masks. We grow a 656-image dataset by more than 400x and validate across two datasets: biopsy-confirmed Diverse Dermatology Images (DDI) and expert-verified Fitzpatrick17k (F17k). On the DDI benchmark, we achieve malignancy classification accuracy of 86.4% under synthetic-only training and 90.9% state-of-the-art performance with real data fine-tuning, alongside leading fairness metrics. Cross-dataset experiments show +13.9% accuracy improvements on unseen F17k data despite minimal disease overlap. We openly release 266k+ synthetic images, code, and generative models to further support fairness research at https://github.com/hectorcarrion/ControllableGenDDI.

Figures

Figures reproduced from arXiv: 2607.12987 by H\'ector Carri\'on, Narges Norouzi.

Figure 1
Figure 1. Figure 1: cgDDI framework. Original images, masks, and prompts produce healthy synthetics. These serve as targets for lesion-mapped synthetics, as prior￾preservation anchors, and as semantic prompts. Disease-specific concepts are learned via textual inversion and used to fine-tune latent diffusion models from which seman￾tic synthetics are sampled. The aggregated data trains fair classifiers. yet Artificial Intellig… view at source ↗
Figure 2
Figure 2. Figure 2: Healthy synthetic imagery. Our inpainting removes target lesions (and markers), producing lesion-less reconstruction robust to hair, morphology, and body location. These samples are later used for lesion mapping and in-distribution prior preservation. Results shown for DDI (human masking) and F17k (algorithmic masking). 3.1 Data Pre-processing DDI contains 656 samples (171 malignant, 485 benign). We consol… view at source ↗
Figure 3
Figure 3. Figure 3: cgDDI samples. Row 1: real images used as prompts. Row 2: lesion from a donor transplanted onto the prompt image. Rows 3–4: semantic synthetics condi￾tioned on the prompt, a target disease (malignant/benign), and target skin tone (light, medium, dark). samples (22 Light, 17 Medium, 16 Dark). We recognize that in ideal conditions these test sets should be larger, and encourage the community to collect more … view at source ↗

Review history (2 revisions) →

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

27 extracted references · 3 canonical work pages · 3 internal anchors

  1. [1]

    arXiv (Cornell University) (2024).https://doi.org/10.48550/arxiv.2409

    Aayushman, Gaddey, H., Mittal, V., Chawla, M., Gupta, G.R.: PatchAlign: Fair and Accurate Skin Disease Image Classification by Alignment with Clinical La- bels. arXiv (Cornell University) (2024).https://doi.org/10.48550/arxiv.2409. 04975

    doi:10.48550/arxiv.2409 2024

  2. [2]

    In: Deep Generative Models

    Akrout, M., Gyepesi, B., Holló, P., et al.: Diffusion-based data augmentation for skin disease classification: Impact across original medical datasets to fully synthetic images. In: Deep Generative Models. pp. 99–109. Springer Nature Switzerland (2024)

    2024

  3. [3]

    Journal of Clinical Oncology27(36), 6199– 6206 (11 2009)

    Balch, C.M., Gershenwald, J.E., Soong, S.J., et al.: Final version of 2009 AJCC Melanoma Staging and Classification. Journal of Clinical Oncology27(36), 6199– 6206 (11 2009)

    2009

  4. [4]

    International Conference on Learning Representations (2018)

    Binkowski, M., Sutherland, D.J., Arbel, M., et al.: Demystifying MMD GANs. International Conference on Learning Representations (2018)

    2018

  5. [5]

    Carion, N., Gustafson, L., Hu, Y.T., Debnath, S., Hu, R., Suris, D., Ryali, C., Alwala, K.V., Khedr, H., Huang, A., Lei, J., Ma, T., Guo, B., Kalla, A., Marks, M., Greer, J., Wang, M., Sun, P., Rädle, R., Afouras, T., Mavroudi, E., Xu, K., Wu, T.H., Zhou, Y., Momeni, L., Hazra, R., Ding, S., Vaze, S., Porcher, F., Li, F., Li, S., Kamath, A., Cheng, H.K., ...

    2025

  6. [6]

    FEDD -- Fair, Efficient, and Diverse Diffusion-based Lesion Segmentation and Malignancy Classification

    Carrión, H., Norouzi, N.: FEDD – Fair, Efficient, and diverse diffusion-based lesion segmentation and malignancy Classification. arXiv (Cornell University) (2023). https://doi.org/10.48550/arxiv.2307.11654

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2307.11654 2023

  7. [7]

    Telemedicine and e-Health25, 1022–1032 (11 2019) 10 Carrión et al

    Coustasse, A., Sarkar, R., Abodunde, B., et al.: Use of teledermatology to improve dermatological access in rural areas. Telemedicine and e-Health25, 1022–1032 (11 2019) 10 Carrión et al

    2019

  8. [8]

    Daneshjou, R., Barata, C., Betz-Stablein, B., et al.: Checklist for evaluation of image-basedartificialintelligencereportsindermatology.JAMADermatology158, 90 (01 2022)

    2022

  9. [9]

    JAMA Dermatology157 (09 2021)

    Daneshjou, R., Smith, M.P., Sun, M.D., et al.: Lack of transparency and potential bias in artificial intelligence data sets and algorithms. JAMA Dermatology157 (09 2021)

    2021

  10. [10]

    Science Advances8(08 2022)

    Daneshjou, R., Vodrahalli, K., Novoa, R.A., et al.: Disparities in dermatology ai performance on a diverse, curated clinical image set. Science Advances8(08 2022)

    2022

  11. [11]

    NIPS (06 2021)

    Dhariwal, P., Nichol, A.: Diffusion models beat gans on image synthesis. NIPS (06 2021)

    2021

  12. [12]

    Lecture Notes in Computer Science13804, 185–202 (2023)

    Du, S., Hers, B., Bayasi, N., et al.: Fairdisco: Fairer ai in dermatology via disentan- glement contrastive learning. Lecture Notes in Computer Science13804, 185–202 (2023)

    2023

  13. [13]

    In: The Eleventh International Conference on Learning Representations (2023)

    Gal, R., Alaluf, Y., Atzmon, Y., et al.: An image is worth one word: Personalizing text-to-image generation using textual inversion. In: The Eleventh International Conference on Learning Representations (2023)

    2023

  14. [14]

    Nature Medicine30(2), 573–583 (2 2024)

    Groh, M., Badri, O., Daneshjou, R., et al.: Deep learning-aided decision support for diagnosis of skin disease across skin tones. Nature Medicine30(2), 573–583 (2 2024)

    2024

  15. [15]

    CVPRW (04 2021)

    Groh, M., Harris, C., Soenksen, L., et al.: Evaluating deep neural networks trained on clinical images in dermatology with the fitzpatrick 17k dataset. CVPRW (04 2021)

    2021

  16. [16]

    Proceedings of the 31st International Conference on Neural Information Processing Systems30, 6626–6637 (2017),https://arxiv.org/pdf/1706.08500

    Heusel, M., Ramsauer, H., Unterthiner, T., et al.: GANs trained by a two Time- Scale update rule converge to a local Nash equilibrium. Proceedings of the 31st International Conference on Neural Information Processing Systems30, 6626–6637 (2017),https://arxiv.org/pdf/1706.08500

    Pith/arXiv arXiv 2017

  17. [17]

    In: International Conference on Learning Representations (2022)

    Hu, E.J., Shen, Y., Wallis, P., et al.: LoRA: Low-rank adaptation of large language models. In: International Conference on Learning Representations (2022)

    2022

  18. [18]

    Nature Medicine30(4), 1166–1173 (4 2024)

    Ktena, I., Wiles, O., Albuquerque, I., et al.: Generative models improve fairness of medical classifiers under distribution shifts. Nature Medicine30(4), 1166–1173 (4 2024)

    2024

  19. [19]

    arXiv (Cornell University) (2023).https://doi.org/10.48550/arxiv.2310.03502

    Razzhigaev, A., Shakhmatov, A., Maltseva, A., Arkhipkin, V., Pavlov, I., Ryabov, I., Kuts, A., Panchenko, A., Kuznetsov, A., Dimitrov, D.: Kandinsky: an Improved Text-to-Image Synthesis with Image Prior and Latent Diffusion. arXiv (Cornell University) (2023).https://doi.org/10.48550/arxiv.2310.03502

    Pith/arXiv arXiv doi:10.48550/arxiv.2310.03502 2023

  20. [20]

    2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) pp

    Rombach, R., Blattmann, A., Lorenz, D., et al.: High-Resolution Image Synthesis with Latent Diffusion Models. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) pp. 10674–10685 (6 2022)

    2022

  21. [21]

    2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) pp

    Ruiz, N., Li, Y., Jampani, V., et al.: DreamBooth: Fine Tuning Text-to-Image diffusion models for Subject-Driven Generation. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) pp. 22500–22510 (6 2023)

    2022

  22. [22]

    arXiv (Cornell Univer- sity) (2023).https://doi.org/10.48550/arxiv.2308.12453

    Sagers, L.W., Diao, J.A., Melas-Kyriazi, L., Groh, M., Rajpurkar, P., Adamson, A.S., Rotemberg, V., Daneshjou, R., Manrai, A.K.: Augmenting medical image classifiers with synthetic data from latent diffusion models. arXiv (Cornell Univer- sity) (2023).https://doi.org/10.48550/arxiv.2308.12453

    Pith/arXiv arXiv doi:10.48550/arxiv.2308.12453 2023

  23. [23]

    In: NeurIPS 2022 Workshop on Synthetic Data for Empowering ML Research (2022)

    Sagers, L.W., Diao, J.A., Groh, M., et al.: Improving dermatology classifiers across populations using images generated by large diffusion models. In: NeurIPS 2022 Workshop on Synthetic Data for Empowering ML Research (2022)

    2022

  24. [24]

    Scientific Data5(08 2018) cgDDI 11

    Tschandl, P., Rosendahl, C., Kittler, H.: The ham10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions. Scientific Data5(08 2018) cgDDI 11

    2018

  25. [25]

    From Majority to Minority: A Diffusion-based Augmentation for Underrepresented Groups in Skin Lesion Analysis

    Wang, J., Chung, Y., Ding, Z., Hamm, J.: From Majority to Minority: A diffusion- based augmentation for underrepresented groups in skin lesion analysis. arXiv (Cornell University) (2024).https://doi.org/10.48550/arxiv.2406.18375

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2406.18375 2024

  26. [26]

    An Improved Method for Personalizing Diffusion Models

    Zeng, Y., Suganuma, M., Okatani, T.: An improved method for personalizing diffusion models. arXiv (Cornell University) (2024).https://doi.org/10.48550/ arxiv.2407.05312

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2407.05312 2024

  27. [27]

    In: CVPR (2018)

    Zhang, R., Isola, P., Efros, A.A., et al.: The unreasonable effectiveness of deep features as a perceptual metric. In: CVPR (2018)

    2018