Pith. sign in

REVIEW 3 major objections 6 minor 26 references

A teacher that trains on lesion masks can transfer their value to a student that never sees them, improving skin-lesion classification while keeping inference image-only.

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

2026-08-01 14:20 UTC pith:APVPWJF7

load-bearing objection A solid experimental protocol and a plausible empirical gain, but the paper's core relational losses are under-specified and the 'lesion-context' partition is shown to be impure, so the mechanism is not established. the 3 major comments →

arxiv 2607.18773 v1 pith:APVPWJF7 submitted 2026-07-21 cs.CV

Privileged Lesion-Context Relational Distillation for Mask-Free Skin Lesion Classification

Abu Mukaddim Rahi , Md Mithun Hossain , Md Zulficar Hasan Joy , M. F. Mridha , Md. Jakir Hossen This is my paper
classification cs.CV
keywords skin lesion classificationknowledge distillationlearning using privileged informationrelational knowledge distillationlesion-context relational learningattention distillationdermoscopic image analysismask-free inference
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 argues that the diagnostic value locked in lesion segmentation masks does not need to be present at inference: a mask-aware teacher can distill that value into an image-only student as relational knowledge. The student and teacher are never asked to match raw feature vectors; instead, both decompose their representations into lesion and context embeddings, and the student learns to reproduce the teacher's similarity geometry, lesion-context affinities, separation, and class-aware organization. In experiments on HAM10000 and external ISIC 2018 (no retraining), this recipe achieves the best balanced accuracy and Macro-F1 among strong image-only and mask-based baselines, with improved calibration and a 0.93M-parameter, 5.5ms student. If the mechanism holds, mask-level supervision can be harvested at training time and deployed at zero extra clinical cost.

Core claim

PLCRD claims that lesion segmentation masks, used only during training, can be converted into a relational representation that an image-only student can inherit without any mask at inference. The student learns to reproduce the teacher's diagnostic distribution, class-conditioned attention, and — the central novelty — the relational geometry between lesion-focused and contextual embeddings. Four loss terms (inter-lesion relation alignment, lesion-context affinity, separation regularization, class-aware relational learning) operate on attention-weighted pooled embeddings z_{q,l} and z_{q,b}, avoiding direct feature matching and enabling heterogeneous teacher-student architectures. Reported re

What carries the argument

The load-bearing object is the pair of embeddings obtained by attention-weighted pooling: z_{q,l} = Pool(F_q, A_q) and z_{q,b} = Pool(F_q, 1−A_q), where A_q is the class-probability-weighted attention map from Eq. (7). A_q is treated as a surrogate lesion mask in both teacher and student, enabling the student to learn a lesion/context decomposition without ever seeing a mask. The four relational losses then enforce that the student's embedding geometry mirrors the teacher's, which is what lets the mask-knowledge survive the teacher's removal.

Load-bearing premise

The premise that the class-probability-weighted attention map A_q is a faithful, transferable proxy for the lesion-versus-context split in both teacher and student — if the student's attention does not actually localize lesions, the relational losses align noise, and the claimed mechanism collapses.

What would settle it

Take the full PLCRD student and re-train it with A_q replaced by a constant (uniform) attention map for the relational losses, keeping logit and attention distillation unchanged. If HAM10000 balanced accuracy and Macro-F1 do not drop materially, the relational transfer of lesion-context structure is not what drives the improvement, and the core claim is falsified.

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

If this is right

  • Any task with training-time masks could adopt the same relational transfer recipe, yielding a mask-free student with no architectural constraints on backbone choice.
  • Because the relational losses act on similarities rather than raw features, the student can be much lighter than the teacher; the reported latency (5.5ms median CPU) supports point-of-care deployment.
  • The calibration gains (ECE 0.022) suggest that mask-informed relational training also corrects confidence estimates, which matters for referral and review workflows.
  • The small internal-to-external drop (balanced accuracy 0.764→0.722) indicates the transferred relational structure generalizes across datasets, at least within similar dermoscopic image distributions.

Where Pith is reading between the lines

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

  • A direct test of the proposed mechanism: replace the student's attention map A_q with the teacher's map or with a fixed random map at inference; if balanced accuracy does not collapse, the lesion-context decomposition is not the carrier of the gain.
  • The same privileged-relational distillation could transfer to other segmentation-to-classification problems (e.g., polyp or lung-nodule imaging) where masks are available during model development but not in the clinic.
  • The reported Soft Dice of 0.342 with high pointing accuracy (0.964) suggests the student attends to compact discriminative sub-regions rather than full lesions; this could be exploited to produce clinically useful, human-interpretable evidence maps.
  • The separation regularizer and the affinity-matching loss pull in opposite directions; their trade-off (β vs α) is likely dataset-sensitive and would need re-tuning when the method moves to new domains.

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

3 major / 6 minor

Summary. The paper proposes PLCRD, a teacher-student framework for skin lesion classification that uses lesion segmentation masks only during training. A mask-aware teacher processes both the original image and a mask-guided lesion view; an image-only student is trained with logit distillation, attention distillation, and a lesion-context relational objective. The student requires no mask at inference. Experiments on HAM10000 (lesion-disjoint five-fold split, one fold held out) and external ISIC 2018 evaluation without retraining report gains in balanced accuracy, Macro-F1, and AUROC over several baselines. The central claim is that privileged mask information can be converted into relational knowledge that transfers to a lightweight image-only classifier.

Significance. If the mechanism and results hold, PLCRD addresses a practical deployment concern: leveraging expensive pixel-level annotations during training while keeping inference mask-free. The experimental design is stronger than typical for this area: lesion-disjoint partitioning, a held-out internal fold, external validation on ISIC 2018 without retraining, and a broad comparison set including segmentation-dependent, localization-guided, and distillation baselines. The paper also reports calibration and efficiency metrics, which strengthens the practical contribution. However, the core relational losses are incompletely specified, and the advertised lesion-context decomposition is called into question by the paper's own localization measurements. These issues must be resolved before the central claim can be accepted.

major comments (3)
  1. [§III-F, Eq. (11)] Eq. (11) defines L_PLCRD as a weighted sum of L_LR, L_LC, L_SEP, and L_CAR, but only L_LR is explicitly defined (Eq. 10). The other three losses are named and qualitatively described but never given equations. Since these losses are the paper's main contribution and their weights α, β, γ are reported, the method is not reproducible without explicit definitions. Provide the exact forms of L_LC, L_SEP, and L_CAR.
  2. [§III-F Eq. (9) and Table VIII] The lesion/context decomposition in Eq. (9) assumes A_q is a valid soft partition separating lesion from context. Table VIII reports Soft Dice = 0.342 for PLCRD's attention, with lesion-attention fraction 0.956. The authors' own interpretation in Sec. VI states the model focuses on compact subregions rather than the complete lesion. Consequently, the complement 1−A_q contains a large portion of the lesion, contaminating the 'context' embedding z_{q,b} with lesion pixels. The relational losses may then align an arbitrary image-dependent partition rather than mask-guided lesion-context structure. The paper provides no diagnostic of partition purity. This directly affects the causal claim in Sec. V-J that lesion-context relational distillation is central to the improvements. Add a quantitative analysis of how much lesion mass falls into the 'context' side, or justify why the compact subregi
  3. [§IV-G, Tables V–VII] Section IV-G states that paired non-parametric tests and bootstrap confidence intervals were performed, but no p-values, confidence intervals, or test outcomes are reported anywhere. Several reported improvements have overlapping standard deviations (e.g., internal balanced accuracy: PLCRD 0.764±0.023 vs. ConvNeXt-Tiny 0.743±0.027; internal Macro-F1: PLCRD 0.773±0.018 vs. Swin-Tiny 0.754±0.006). Table VII explicitly labels its differences as 'descriptive effect estimates,' while Sec. V-C and Sec. V-J present them as improvements. Without significance evidence, the claim that PLCRD outperforms baselines is not statistically supported. Report the planned statistical comparisons.
minor comments (6)
  1. [Author affiliation] Typo: 'BUBT Resarch Graduate School' should be 'Research'.
  2. [Section I, first paragraph] The text begins with 'S KIN lesion classification' — likely a formatting artifact from the source; fix spacing.
  3. [Eq. (4)] The symbol e m_i is used but not defined; presumably it is the resized binary mask. Please define it in the text.
  4. [Section V-D, Table VIII] The metric 'Soft Dice' is not defined. Clarify how it is computed for continuous attention maps against binary masks.
  5. [Table VI] The check-mark encoding for the PLCRD row appears to indicate that PLCRD uses Spatial Relation, but the text describes spatial relation distillation as a separate baseline. Clarify whether the relational component in PLCRD is the same as 'Spatial Relation' or different.
  6. [Figure 5] The caption mentions a 'lesion-context partition' but does not explain how the partition is generated for visualization. Clarify the relationship between the attention map and the partition shown.

Circularity Check

1 steps flagged

Classification claims are externally benchmarked; only Fig. 5's relation-matrix agreement is self-referential because it is the L_LR training objective.

specific steps
  1. fitted input called prediction [Sec. III-F (Eqs. 10-11) and Sec. V-G / Fig. 5]
    "The student is trained to match the teacher relation matrix, allowing the student to inherit the teacher’s mask-informed lesion geometry. ... Cosine similarity, correlation, and normalized Frobenius error are used to quantify agreement between the student relation matrices and the privileged teacher."

    The agreement visualized in Fig. 5 is the quantity directly minimized by L_LR: the student relation matrix R^{S,l} is optimized to match R^{T,l} (Eqs. 10-11). Reporting high cosine/correlation/low Frobenius agreement between those same matrices is therefore reporting the training loss itself, not an independent confirmation of relational transfer. This is self-referential but not load-bearing for the main external classification result.

full rationale

No significant circularity in the central claim. The reported classification gains (Sec. V-A) are measured on a held-out HAM10000 test fold and on ISIC 2018 without retraining, against image-only, ROI, segmentation-guided, and distillation baselines; the student is not evaluated against the teacher. Hyperparameters are set by validation and the teacher is frozen, so the student's test performance is not an optimized fit to the test set. The paper contains no load-bearing self-citations or uniqueness theorems: the references are standard external works. The only self-referential element is Sec. V-G / Fig. 5, where student-teacher relation-matrix agreement is presented as evidence even though L_LR explicitly optimizes that agreement; this is a minor interpretability overstatement, not a derivation that reduces the main result to its input. The concern about A_q/1-A_q partition purity (Table VIII Soft Dice 0.342) is a correctness/mechanism question, not circularity.

Axiom & Free-Parameter Ledger

7 free parameters · 5 axioms · 0 invented entities

No new physical or ontological entities are introduced. The method rests on hand-chosen loss weights, attention-as-lesion-proxy, and dataset assumptions about mask quality, lesion grouping, and external labels.

free parameters (7)
  • logit KD weight λ_KD = 0.75
    Eq. (12) combines student objectives; the value was chosen by hand/validation, not derived.
  • attention distillation weight λ_att = 0.10
    Eq. (12); chosen by hand/validation.
  • PLCRD weight λ_PLCRD = 0.10
    Eq. (12); chosen by hand/validation.
  • relational component weights α, β, γ = 0.25, 0.10, 0.25
    Eq. (11) balances inter-lesion alignment, lesion-context affinity/separation, and class-aware relation; fixed by hand/validation.
  • distillation temperature τ = 4
    Eq. (6); chosen by hand/validation.
  • minimum mask lesion-pixel threshold = 64
    Table III: masks with fewer than 64 lesion pixels were excluded from mask-based supervision; an ad hoc preprocessing choice.
  • class-balancing weights w_yi = not reported
    Eq. (5) uses class-balancing weights to handle imbalance, but the exact values are not given.
axioms (5)
  • domain assumption Ground-truth lesion masks in HAM10000 correctly delineate lesion regions for constructing the privileged teacher and localization supervision.
    Eq. (2) builds x_l = x ⊙ m and Eq. (4) uses masks to supervise teacher attention; if masks are noisy, the teacher learns from incorrect regions.
  • domain assumption Lesion-disjoint metadata in HAM10000 correctly groups all images of the same lesion, so the fixed fold-0 test has no near-duplicate leakage.
    Section IV-B describes lesion-level partitioning to prevent leakage, but the correctness of the lesion-identity metadata is assumed.
  • ad hoc to paper The class-conditioned attention map A_q is a valid soft partition of the image into lesion versus context for both teacher and student.
    Eq. (9) defines z_q,l and z_q,b by pooling with A_q and 1−A_q; if attention is not a faithful lesion proxy, the relational objectives align meaningless regions.
  • ad hoc to paper Relational structure of teacher embeddings, rather than raw teacher features, is sufficient to transfer mask-informed knowledge to a heterogeneous student.
    This is the premise of Section III-F; no proof is given, only empirical ablation support.
  • domain assumption ISIC 2018 test set is a valid external distribution and its labels are reliable; excluding image ISIC_0035068 does not bias results.
    Section IV-A reports external validation without retraining; label reliability and the effect of the single excluded image are assumed.

pith-pipeline@v1.3.0-alltime-deepseek · 16964 in / 14505 out tokens · 133072 ms · 2026-08-01T14:20:56.615258+00:00 · methodology

0 comments
read the original abstract

Accurate skin lesion classification can benefit from lesion segmentation masks, but requiring masks or an auxiliary segmentation model during inference reduces clinical practicality and increases computational complexity. This work introduces Privileged Lesion-Context Relational Distillation (PLCRD), a teacher-student framework that exploits lesion masks exclusively during training while preserving image-only inference. The privileged teacher jointly analyzes the original dermoscopic image and its mask-guided lesion region to learn lesion-specific and contextual diagnostic representations. An image-only student is then trained through complementary knowledge-transfer mechanisms that convey the teacher's diagnostic distribution, lesion-focused attention, inter-lesion relational geometry, and lesion-context structure. PLCRD decomposes deep representations into lesion and contextual embeddings and transfers their relational organization through inter-lesion similarity alignment, lesion-context affinity matching, separation regularization, and class-aware relational learning. This formulation avoids direct feature matching between heterogeneous teacher and student architectures and enables the student to internalize mask-informed diagnostic structure without accessing masks at deployment. The framework was evaluated on HAM10000 using lesion-disjoint data partitioning and externally validated on ISIC 2018 without retraining. PLCRD achieved a lesion-level macro-F1 of 0.773 +/- 0.018, balanced accuracy of 0.764 +/- 0.023, and macro-AUROC of 0.976 +/- 0.002 on HAM10000, together with a macro-F1 of 0.732 +/- 0.008 on ISIC 2018. The results indicate that privileged lesion annotations can be transformed into transferable relational knowledge, yielding a practical and interpretable approach to mask-free skin lesion classification.

Figures

Figures reproduced from arXiv: 2607.18773 by Abu Mukaddim Rahi, Md. Jakir Hossen, Md Mithun Hossain, Md Zulficar Hasan Joy, M. F. Mridha.

Figure 1
Figure 1. Figure 1: Overview of the proposed PLCRD framework. The mask-aware teacher uses lesion masks as privileged information [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Representative dermoscopic samples from HAM10000 and their corresponding lesion segmentation masks. The first [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Qualitative comparison of class-discriminative Grad-CAM++ visualizations for representative internal and external skin [PITH_FULL_IMAGE:figures/full_fig_p010_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Case-based comparison of the image-only student, PMD baseline, and PLCRD. Panel (a) shows an image-only error [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Visualization of lesion-context relational structure for two representative benign keratosis-like lesion cases. For each [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗

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

26 extracted references · 8 linked inside Pith

  1. [1]

    Dermatologist-level classification of skin cancer with deep neural networks,

    A. Esteva, B. Kuprel, R. A. Novoa, J. Ko, S. M. Swetter, H. M. Blau, and S. Thrun, “Dermatologist-level classification of skin cancer with deep neural networks,”Nature, vol. 542, no. 7639, pp. 115–118, 2017

    2017

  2. [2]

    Man against machine: Diagnostic performance of a deep learning convolutional neural network for dermoscopic melanoma recognition in comparison to 58 dermatologists,

    H. A. Haenssle, C. Fink, R. Schneiderbauer, F. Toberer, T. Buhl, A. Blum, A. N. Kalloo, A. Ben Hadj Hassen, L. Thomas, A. Enk, L. Uhlmann, C. Alt, M. Arenbergerova, R. Bakos, E. Baltas, I. Bertlich, A. Blumet al., “Man against machine: Diagnostic performance of a deep learning convolutional neural network for dermoscopic melanoma recognition in comparison...

    2018

  3. [3]

    Comparison of the accuracy of human readers versus machine-learning algorithms for pigmented skin lesion classification: An open, web-based, international, diagnostic study,

    P. Tschandl, N. Codella, B. N. Akay, G. Argenziano, R. P. Braun, H. Cabo, D. Gutman, A. Halpern, B. Helba, R. Hofmann-Wellenhof, H. Kittler, M. A. Marchettiet al., “Comparison of the accuracy of human readers versus machine-learning algorithms for pigmented skin lesion classification: An open, web-based, international, diagnostic study,”The Lancet Oncolog...

    2019

  4. [4]

    The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions,

    P. Tschandl, C. Rosendahl, and H. Kittler, “The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions,”Scientific Data, vol. 5, p. 180161, 2018. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 14

    2018

  5. [5]

    Skin lesion analysis toward melanoma detection 2018: A challenge hosted by the international skin imaging collaboration (ISIC),

    N. Codella, V . Rotemberg, P. Tschandl, M. E. Celebi, S. Dusza, D. Gut- man, B. Helba, A. Kalloo, K. Liopyris, M. Marchetti, H. Kittler, and A. Halpern, “Skin lesion analysis toward melanoma detection 2018: A challenge hosted by the international skin imaging collaboration (ISIC),” arXiv preprint arXiv:1902.03368, 2019

    Pith/arXiv arXiv 2018

  6. [6]

    The effects of skin lesion segmentation on the performance of dermatoscopic image classification,

    A. Mahbod, P. Tschandl, G. Langs, R. Ecker, and I. Ellinger, “The effects of skin lesion segmentation on the performance of dermatoscopic image classification,”Computer Methods and Programs in Biomedicine, vol. 197, p. 105725, 2020

    2020

  7. [7]

    A new learning paradigm: Learning using privileged information,

    V . Vapnik and A. Vashist, “A new learning paradigm: Learning using privileged information,”Neural Networks, vol. 22, no. 5–6, pp. 544–557, 2009

    2009

  8. [8]

    Distilling the knowledge in a neural network,

    G. Hinton, O. Vinyals, and J. Dean, “Distilling the knowledge in a neural network,”arXiv preprint arXiv:1503.02531, 2015

    Pith/arXiv arXiv 2015

  9. [9]

    Paying more attention to attention: Improving the performance of convolutional neural networks via atten- tion transfer,

    S. Zagoruyko and N. Komodakis, “Paying more attention to attention: Improving the performance of convolutional neural networks via atten- tion transfer,” inInternational Conference on Learning Representations, 2017

    2017

  10. [10]

    M. A. Marchetti, N. C. F. Codella, S. W. Dusza, D. A. Gutman, B. Helba, A. Kalloo, K. Liopyris, N. Mishra, H. Kittler, and A. C. Halpern, “Results of the 2016 international skin imaging collaboration interna- tional symposium on biomedical imaging challenge: Comparison of the accuracy of computer algorithms to dermatologists for the diagnosis of melanoma ...

    2016

  11. [11]

    Expert-level diagnosis of nonpigmented skin cancer by combined convolutional neural networks,

    P. Tschandl, C. Rosendahl, B. N. Akay, G. Argenziano, A. Blum, R. P. Braun, H. Cabo, J.-Y . Gourhant, J. Kreusch, A. Lallas, J. Malvehy et al., “Expert-level diagnosis of nonpigmented skin cancer by combined convolutional neural networks,”JAMA Dermatology, vol. 155, no. 1, pp. 58–65, 2019

    2019

  12. [12]

    Acral melanoma detection using a convolutional neural network for dermoscopy images,

    C. Yu, S. Yang, W. Kim, J. Jung, K. H. Chung, and S. W. Lee, “Acral melanoma detection using a convolutional neural network for dermoscopy images,”PLOS ONE, vol. 13, no. 3, p. e0193321, 2018

    2018

  13. [13]

    Gutman, N

    D. Gutman, N. C. F. Codella, M. E. Celebi, B. Helba, M. Marchetti, N. Mishra, and A. Halpern, “Skin lesion analysis toward melanoma detection: A challenge at the international symposium on biomedical imaging (isbi) 2016, hosted by the international skin imaging collabo- ration (isic),”arXiv preprint arXiv:1605.01397, 2016

    Pith/arXiv arXiv 2016

  14. [14]

    N. C. F. Codella, D. Gutman, M. E. Celebi, B. Helba, M. A. Marchetti, S. W. Dusza, A. Kalloo, K. Liopyris, N. Mishra, H. Kittler, and A. Halpern, “Skin lesion analysis toward melanoma detection: A chal- lenge at the 2017 international symposium on biomedical imaging, hosted by the international skin imaging collaboration (isic),”arXiv preprint arXiv:1710....

    Pith/arXiv arXiv 2017

  15. [15]

    Skin lesion classification using hybrid deep neural networks,

    A. Mahbod, G. Schaefer, C. Wang, R. Ecker, and I. Ellinger, “Skin lesion classification using hybrid deep neural networks,”arXiv preprint arXiv:1702.08434, 2017

    Pith/arXiv arXiv 2017

  16. [16]

    A multi-level deep ensemble model for skin lesion classification in dermoscopy images,

    Y . Xie, J. Zhang, and Y . Xia, “A multi-level deep ensemble model for skin lesion classification in dermoscopy images,”arXiv preprint arXiv:1807.08488, 2018

    Pith/arXiv arXiv 2018

  17. [17]

    Efficientnet: Rethinking model scaling for convolutional neural networks,

    M. Tan and Q. V . Le, “Efficientnet: Rethinking model scaling for convolutional neural networks,” inProceedings of the 36th International Conference on Machine Learning, 2019, pp. 6105–6114

    2019

  18. [18]

    Searching for mobilenetv3,

    A. Howard, M. Sandler, G. Chu, L.-C. Chen, B. Chen, M. Tan, W. Wang, Y . Zhu, R. Pang, V . Vasudevan, Q. V . Le, and H. Adam, “Searching for mobilenetv3,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2019, pp. 1314–1324

    2019

  19. [19]

    Swin transformer: Hierarchical vision transformer using shifted windows,

    Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, and B. Guo, “Swin transformer: Hierarchical vision transformer using shifted windows,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2021, pp. 10 012–10 022

    2021

  20. [20]

    Rethinking vision transformers for mobilenet size and speed,

    Y . Li, J. Hu, Y . Wen, G. Evangelidis, K. Salahi, Y . Wang, S. Tulyakov, and J. Ren, “Rethinking vision transformers for mobilenet size and speed,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 17 289–17 300

    2023

  21. [21]

    Segmentation and classification of skin lesions for disease diagnosis,

    R. Sumithra, M. M. Suhil, and D. S. Guru, “Segmentation and classification of skin lesions for disease diagnosis,”arXiv preprint arXiv:1609.03277, 2016

    Pith/arXiv arXiv 2016

  22. [22]

    Fitnets: Hints for thin deep nets,

    A. Romero, N. Ballas, S. E. Kahou, A. Chassang, C. Gatta, and Y . Ben- gio, “Fitnets: Hints for thin deep nets,” inInternational Conference on Learning Representations, 2015

    2015

  23. [23]

    Multiple skin lesions diagnostics via integrated deep convolutional networks for segmentation and classification,

    M. A. Al-Masni, D.-H. Kim, and T.-S. Kim, “Multiple skin lesions diagnostics via integrated deep convolutional networks for segmentation and classification,”Computer methods and programs in biomedicine, vol. 190, p. 105351, 2020

    2020

  24. [24]

    A novel multi-task deep learning model for skin lesion segmentation and classification,

    X. Yang, Z. Zeng, S. Y . Yeo, C. Tan, H. L. Tey, and Y . Su, “A novel multi-task deep learning model for skin lesion segmentation and classification,”arXiv preprint arXiv:1703.01025, 2017

    Pith/arXiv arXiv 2017

  25. [25]

    An end-to-end multi- task deep learning framework for skin lesion analysis,

    L. Song, J. Lin, Z. J. Wang, and H. Wang, “An end-to-end multi- task deep learning framework for skin lesion analysis,”IEEE journal of biomedical and health informatics, vol. 24, no. 10, pp. 2912–2921, 2020

    2020

  26. [26]

    Relational knowledge distilla- tion,

    W. Park, D. Kim, Y . Lu, and M. Cho, “Relational knowledge distilla- tion,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 3967–3976

    2019