Pith. sign in

REVIEW 6 major objections 6 minor 31 references

Optimizing Deep Learning for Skin Cancer Classification: A Computationally Efficient CNN with Minimal Accuracy Trade-Off

T0 review · 6 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A custom CNN with 692,807 parameters classifies skin lesions on HAM10000 within the reported accuracy margin of a ResNet50 transfer-learning baseline while using far fewer FLOPs.

desk verdict The paper's central efficiency claim collapses under its own architecture table; desk reject unless the authors redo the arithmetic and release code. read the letter →

arxiv 2505.21597 v1 pith:UKZU66AD submitted 2025-05-27 eess.IV cs.CVcs.LG

Abdullah Al Mamun , Pollob Chandra Ray , Md Rahat Ul Nasib , Akash Das , Jia Uddin , Md Nurul Absur This is my paper
classification eess.IVcs.CVcs.LG
keywords skincancerclassificationconvolutionalneuralnetworktransferlearningHAM10000FLOPreductionparameterefficiencymelanomadetectionedgedeployment
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

This paper argues that a deliberately small convolutional network can give nearly the same skin-lesion classification accuracy as a ResNet50 transfer-learning baseline while consuming only a tiny fraction of the compute. On the HAM10000 dataset it reports 87.05% accuracy for the custom CNN versus 89.08% for ResNet50 with transfer learning, a difference the paper describes as less than 0.022%. The efficiency claims are a 96.7% parameter reduction (692,807 versus 23.9 million) and a 99.2% FLOP reduction (30.04 million versus 4.00 billion). The authors' point is that for deployment on mobile and edge devices, this trade-off may be preferable to a heavyweight model.

What carries the argument

The central object is the proposed custom CNN architecture of Table II: three Conv2D-MaxPool blocks with 32, 64, and 128 filters using $3\times3$ kernels, followed by a Flatten layer, a 256-unit Dense layer with 50% dropout, and a 7-unit softmax output. The measurement machinery is the pair of FLOP formulas in the paper, one for convolutional layers and one for dense layers, which the authors use to compare the custom network against the ResNet50 transfer-learning baseline. The residual connections in ResNet50 are explained as the contrast: they preserve accuracy but add the computational cost the paper wants to avoid.

What would settle it

Sum the parameters of each layer in Table II using standard formulas and apply the paper's FLOP equations (10) and (11) to the same layer shapes; if the resulting totals are not 692,807 and 30.04 million, the headline reductions do not describe the printed architecture. Recomputing the accuracy difference from the published confusion matrix would similarly settle whether the gap is below 0.022%.

Watch

Extended reading notes

Core claim

The central discovery, as the authors present it, is that a purpose-built CNN with only one fully connected hidden layer can match most of the accuracy of a transfer-learned ResNet50 on the seven-class HAM10000 lesion dataset. The paper reports 87.05% accuracy and 30.04 million FLOPs for the custom network, versus 89.08% accuracy and 4.00 billion FLOPs for ResNet50, a gap it states as less than 0.022%. It also reports a 96.7% parameter reduction and over 60% less training time, and concludes that such lightweight models are viable for mobile and edge diagnostics.

Load-bearing premise

The central comparison rests on the totals in Tables III and IV being measured consistently for the same models: 692,807 parameters, 30.04 million FLOPs, and a reported accuracy gap of less than 0.022% for the custom CNN; if those totals do not correspond to the described networks, the efficiency and minimal-trade-off claims are not anchored.

Editorial extensions

If this is right

  • Devices with tight memory and power budgets could run per-image screening at 30.04 million FLOPs while staying close to the baseline's reported accuracy.
  • The reported 96.7% parameter cut and 99.2% FLOP cut would shrink model size, energy per inference, and latency, which are the constraints that block edge deployment.
  • A clinic that values throughput over a fraction of a percent of accuracy could use the custom CNN as the primary model and reserve ResNet50 for ambiguous cases.
  • The authors' planned quantization and pruning could drive computational cost down further, making the trade-off even more favorable for embedded hardware.
  • The comparison is specific to the seven HAM10000 classes, so applying the same architecture to a different lesion taxonomy would require retraining and re-validation.

Reading between the lines

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

  • Because the paper compares against a single transfer-learning baseline, its conclusion is specific to HAM10000; a multi-site dermoscopy benchmark would test whether the reported accuracy gap persists with more lesion diversity.
  • A from-scratch version of the custom CNN on the same data would isolate whether the gap comes from model capacity or from ImageNet initialization, a distinction the paper does not examine.
  • The percentage reductions depend on which ResNet50 head is used; comparing against a standard pretrained ResNet50 with no custom head would show how much of the saving is architectural rather than a transfer-learning choice.
  • Recomputing the accuracy difference as a relative change rather than an absolute gap would clarify whether the 0.022% wording is a rounding convention or a different metric than the table values suggest.
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

6 major / 6 minor

Summary. The paper proposes a custom CNN for seven-class skin-lesion classification on HAM10000 and compares it with ResNet50-based transfer learning. The central claim is that the custom CNN reduces parameters from 23.9 million to 692,807 (a 96.7% reduction) and FLOPs from 4.00 billion to 30.04 million (a 99.2% reduction) while staying within 0.022% accuracy of the transfer-learning baseline. These figures appear in the Abstract, Section V, Table III, and Table IV. The manuscript also claims a 60% training-time reduction and a 48.5% inference speedup, and asserts that the ResNet50 baseline achieves 89.08% accuracy versus the custom CNN's 87.05%.

Significance. The practical goal of a lightweight, deployable skin-cancer classifier is timely, and the emphasis on FLOPs as a comparison metric is appropriate for edge-deployment arguments. If the reported efficiency figures were reproducible, the paper would provide a useful benchmark. However, the central quantitative claims cannot be reconstructed from the paper's own architecture description and results, so the significance of the contribution as written is not established. The manuscript also provides no code, training hyperparameters, data split, or timing methodology, which further limits verification of the efficiency claims.

major comments (6)
  1. [Table II and Abstract/Conclusion] The parameter total claimed for the custom CNN is not supported by the architecture table. Table II lists five parameter-bearing layers: Conv2D with 896, Conv2D with 18,496, Conv2D with 73,856, Dense with 25,690,112, and Dense with 1,799 parameters. These sum to 25,785,159 parameters, not 692,807. Consequently, the 96.7% parameter reduction claimed in the Abstract, Section V, and Table III cannot be derived from the described architecture.
  2. [Section III-G Eqs. (10)-(11) and Table II] The reported 30.04 million FLOPs is inconsistent with the paper's own FLOP formulas applied to Table II. Using Eq. (10) for the three convolutional layers gives 43.35M, 231.21M, and 231.21M FLOPs, and using Eq. (11) for the two dense layers gives 25.69M and 0.002M FLOPs, for a total of approximately 531.5M FLOPs. This is about 17.7 times the stated 30.04M and already uses the paper's favorable single-multiply counting. The 99.2% FLOP reduction claim therefore cannot be reproduced from the reported architecture.
  3. [Section IV.B and Table III] Section IV.B contradicts Table III and the Abstract by swapping the labels of the two models. The text states that 'ResNet50 + TL lowers the total parameters to 692,807' while 'our custom CNN ... includes just 23.66 million parameters,' but Table III and the Abstract assign 23,661,703 parameters to ResNet50+TL and 692,807 to the custom CNN. This is an internal inconsistency in a load-bearing comparison, and the reader cannot determine which assignment is correct.
  4. [Table IV and Abstract] The claimed 'accuracy deviation of less than 0.022%' is arithmetically wrong. Table IV reports accuracies of 87.05% and 89.08%, whose difference is 2.03 percentage points, not 0.022 percentage points. The Abstract and Section IV.C repeat the 0.022% figure, understating the accuracy gap by roughly two orders of magnitude. This directly undermines the asserted 'minimal accuracy trade-off.'
  5. [Section III.D and Table II] The paper states that binary cross-entropy loss is used with labels 0/1 for benign versus malignant, but Table II specifies a seven-unit output with softmax and the dataset is described as having seven lesion classes. The training objective is therefore not specified consistently, which makes the reported accuracies difficult to interpret. This should be resolved by stating whether a seven-class categorical loss or a binarized melanoma-versus-rest task was used.
  6. [Section V] The Conclusion claims 'over a 60% reduction in training time and a 48.5% speedup in inference,' but no timing experiments, hardware description, or measurement protocol appear anywhere in the manuscript. This is an additional efficiency claim that is load-bearing for the 'computationally efficient' characterization and is currently unsupported.
minor comments (6)
  1. [Section III.E Eq. (5)] The Adam update rule is written incorrectly: the denominator should be the square root of the second-moment estimate plus epsilon, and the numerator should be the first-moment estimate, not v_t divided by sqrt(m_t).
  2. [Section II and Table IV] The paper says the proposed system has an accuracy of 87.08% in Section II, but Table IV reports 87.05%. This minor inconsistency should be corrected.
  3. [Section III-B and Reference [30]] Reference [30] is cited to support 'sequential correlations' as a data-augmentation technique, but the cited paper appears to be about order dependency in sequence correlation and does not obviously describe image augmentation. The authors should remove or justify this citation.
  4. [Section IV] The accuracy and loss curves, confusion matrix, and ROC curves are presented only for the ResNet50 transfer-learning model; no training curves or confusion matrix are shown for the proposed custom CNN, so the reader cannot verify its generalization behavior.
  5. [Section III-G] The terminology 'FLOP' and 'FLOPs' is used inconsistently; the text defines FLOPs as 'Floating Point Operations per Second,' but the tables and equations use FLOPs to mean the total number of floating-point operations. This should be unified.
  6. [Table I] The column labeled 'Param. Red.' lists 'Yes' for the proposed method without a value or a definition of how parameter reduction is measured; the comparison would be clearer if the actual reduction percentage or parameter count were given.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: FLOPs and parameters are direct computations from the stated architecture and the accuracies are empirical; the sole self-citation [30] is peripheral and not load-bearing. Internal arithmetic inconsistencies are correctness defects, not circularity.

full rationale

The paper's central claims—692,807 parameters and 30.04M FLOPs for the custom CNN versus 23.66M parameters and 4.00B FLOPs for the transfer-learned ResNet50, plus the two empirical accuracies—are computed or measured, not derived from a fitted input. The FLOP figures are direct applications of the paper's own Eqs. (10)-(11) to the architecture in Table II; accuracy and loss are empirical test-set results on HAM10000. There is no parameter fitted to a subset of data and then renamed a prediction, no quantity defined in terms of the quantity it is said to predict, and no uniqueness theorem imported from the authors' prior work. The only self-citations are [30], used to justify 'sequential correlations' in data augmentation in Section III.B, and peripheral co-author works [2] and [6] cited in the Introduction; none is load-bearing for the efficiency or accuracy results, which are self-contained empirical comparisons. The important caveat is that the paper contains serious internal arithmetic inconsistencies—Table II's parameter sum (~25.8M) does not reproduce the 692,807 attributed to the custom CNN, Section IV.B swaps the parameter labels between the two models, and the reported 0.022% accuracy deviation contradicts the 89.08% versus 87.05% figures in Table IV. These are correctness and reproducibility defects, not circularity, since a wrong arithmetic result is still a direct computation rather than an input recycled as an output. Per the review rules, internal contradiction is a correctness risk and does not raise the circularity score. Verdict: no significant circularity; score 2 reflects only the presence of a minor, non-load-bearing self-citation [30].

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

No new physical entities or mediators are introduced. The free parameters are the hand-chosen architecture and training choices, which are not justified by ablation. The key unstated premise is that the layer counts in Table II correspond to the total counts in Table III, and that premise fails numerically.

free parameters (4)
  • Custom CNN layer filter counts and dense width = 32, 64, 128, 256
    Filter counts and dense width are chosen without ablation or stated design rationale (Table II).
  • Dropout rate = 0.5
    Dropout rate is set to 0.5 without sensitivity analysis (Table II).
  • Input image resolution = 224x224
    Input size is fixed to match ResNet50 preprocessing; no study of resolution impact is reported (Section III-B).
  • Adam optimizer hyperparameters = not stated
    Learning rate, momentum parameters, and schedule are not reported (Section III-E).
assumptions (4)
  • domain assumption HAM10000 labels are reliable ground truth for seven skin lesion classes.
    The paper treats dataset labels as correct without lesion-level review or expert verification (Section III-B).
  • domain assumption ImageNet-pretrained weights transfer useful features to dermoscopic images.
    Transfer learning is justified by pretraining on ImageNet, but the domain shift from natural images to dermoscopy is not analyzed (Section III-C).
  • standard math The FLOP formulas in Eq. (10) and (11) correctly measure computational cost for the reported layers.
    The formulas are standard, but applying them to the paper's own Table II architecture does not yield the reported 30.04M FLOPs (Section III-G).
  • standard math Residual connection formula y = F(x, {W_i}) + x describes the ResNet50 baseline.
    Eq. (2) is the standard residual block definition and is not the source of the paper's numerical problems.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Optimizing Deep Learning for Skin Cancer Classification: A Computationally Efficient CNN with Minimal Accuracy Trade-Off." pith.science (2026). https://pith.science/paper/UKZU66AD

@misc{pith2026250521597,
  author       = {Pith},
  title        = {Pith review of: Optimizing Deep Learning for Skin Cancer Classification: A Computationally Efficient CNN with Minimal Accuracy Trade-Off},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UKZU66AD}},
  note         = {Machine review of arXiv:2505.21597}
}
read the original abstract

The rapid advancement of deep learning in medical image analysis has greatly enhanced the accuracy of skin cancer classification. However, current state-of-the-art models, especially those based on transfer learning like ResNet50, come with significant computational overhead, rendering them impractical for deployment in resource-constrained environments. This study proposes a custom CNN model that achieves a 96.7\% reduction in parameters (from 23.9 million in ResNet50 to 692,000) while maintaining a classification accuracy deviation of less than 0.022\%. Our empirical analysis of the HAM10000 dataset reveals that although transfer learning models provide a marginal accuracy improvement of approximately 0.022\%, they result in a staggering 13,216.76\% increase in FLOPs, considerably raising computational costs and inference latency. In contrast, our lightweight CNN architecture, which encompasses only 30.04 million FLOPs compared to ResNet50's 4.00 billion, significantly reduces energy consumption, memory footprint, and inference time. These findings underscore the trade-off between the complexity of deep models and their real-world feasibility, positioning our optimized CNN as a practical solution for mobile and edge-based skin cancer diagnostics.

Figures

Figures reproduced from arXiv: 2505.21597 by the authors.

Figure 1
Figure 1. Proposed System Methodology [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Proposed Model [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Classes of Skin Type TABLE III: Comparison of Total Parameters in Different Models. Model Total Parameters MobileNet 4,253,864 Soft Attention 47,535,287 VGG16 + SA 62,650,711 ResNet50 + SA 23,911,319 ResNet34 + SA 31,414,167 IRV2 + SA 47,535,287 DenseNet201 + SA 17,477,463 IRV2 + GradCam 47,550,631 Baseline: ResNet50 + TL 23,661,703 Custom CNN Implementation 692,807 C. Comparative FLOP Analysis Table IV presents a F… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Model Accuracy Curve [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Model Loss Curve E. Confusion Matrix Analysis The confusion matrix for the ResNet50 model’s classification of skin cancer across seven classes reveals a high accuracy in per￾formance, with only minor misclassifications observed, particularly between closely related ski…
Figure 7
Figure 7. Figure 7: ROC Curve contrast, our proposed convolutional neural network (CNN) signifi￾cantly reduces the total parameters from 23.9 million to 692,000—a remarkable 96.7% reduction—and decreases the FLOPs from 4.00 billion to 30.04 million, achieving a 99.2% reduction, all while …

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

31 extracted references · 30 canonical work pages

  1. [30]

    Order depen- dency in sequential correlation,

    K. F. Ahmed Nasif, M. Nurul Absur, and M. Al Mamun, “Order depen- dency in sequential correlation,” in 2019 3rd International Conference on Electrical, Computer & Telecommunication Engineering (ICECTE), 2019, pp. 49–52

    work page 2019

  2. [1]

    S. N. Nova, M. S. Rahman, and C. Chakraborty, Patients’ Health Surveillance Model Using IoT and 6G Technology . Cham: Springer International Publishing, 2021, pp. 191–209. [Online]. Available: https://doi.org/10.1007/978-3-030-73295-0_9

    work page doi:10.1007/978-3-030-73295-0_9 2021

  3. [2]

    Revolutionizing image recognition: Next-generation cnn architectures for handwritten digits and objects,

    M. N. Absur, K. F. A. Nasif, S. Saha, and S. N. Nova, “Revolutionizing image recognition: Next-generation cnn architectures for handwritten digits and objects,” in 2024 IEEE Symposium on Wireless Technology & Applications (ISWTA) , 2024, pp. 173–178

    work page 2024

  4. [3]

    Healthcare professionals creden- tial verification model using blockchain-based self-sovereign identity,

    S. Saha, S. N. Nova, and M. I. Iqbal, “Healthcare professionals creden- tial verification model using blockchain-based self-sovereign identity,” in Proceedings of the Fourth International Conference on Trends in Computational and Cognitive Engineering , M. S. Kaiser, S. Waheed, A. Bandyopadhyay, M. Mahmud, and K. Ray, Eds. Singapore: Springer Nature Singa...

    work page 2023

  5. [4]

    Highly sensitive terahertz metasurface based on electromagnetically induced transparency-like resonance in detection of skin cancer cells,

    S. Nourinovin, M. M. Rahman, M. Naftaly, M. P. Philpott, Q. H. Abbasi, and A. Alomainy, “Highly sensitive terahertz metasurface based on electromagnetically induced transparency-like resonance in detection of skin cancer cells,” IEEE Transactions on Biomedical Engineering, vol. 71, no. 7, pp. 2180–2188, 2024

    work page 2024

  6. [5]

    Melanoma management: From epidemiology to treatment and latest advances,

    J. Lopes, C. M. P. Rodrigues, M. M. Gaspar, and C. P. Reis, “Melanoma management: From epidemiology to treatment and latest advances,” Cancers, vol. 14, no. 19, 2022. [Online]. Available: https://www.mdpi.com/2072-6694/14/19/4652

    work page 2022

  7. [6]

    Anomaly detection in biomedical data and image using various shallow and deep learning algorithms,

    M. N. Absur, “Anomaly detection in biomedical data and image using various shallow and deep learning algorithms,” in Data Intelligence and Cognitive Informatics, I. J. Jacob, S. Kolandapalayam Shanmugam, and R. Bestak, Eds. Singapore: Springer Nature Singapore, 2022, pp. 45– 58

    work page 2022

  8. [7]

    S. N. Nova, M. S. Rahman, and A. S. M. S. Hosen, Deep Learning in Biomedical Devices: Perspectives, Applications, and Challenges . Singapore: Springer Nature Singapore, 2022, pp. 13–35. [Online]. Available: https://doi.org/10.1007/978-981-19-4189-4_2

    work page doi:10.1007/978-981-19-4189-4_2 2022

Show all 31 references
  1. [8]

    Gan-based data augmentation and anonymization for skin-lesion analysis: A critical review,

    A. Bissoto, E. Valle, and S. Avila, “Gan-based data augmentation and anonymization for skin-lesion analysis: A critical review,” in 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2021, pp. 1847–1856

    2021

  2. [9]

    Deep learning-based system for automatic melanoma detection,

    A. A. Adegun and S. Viriri, “Deep learning-based system for automatic melanoma detection,” IEEE Access, vol. 8, pp. 7160–7172, 2020

    2020

  3. [10]

    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,” Sci. Data, vol. 5, no. 1, p. 180161, Aug. 2018

    2018

  4. [11]

    Innovative way of identifying skin cancer model design with fcnn and lstm,

    A. J. Rao, M. Babu, G. DurgaRao, G. P. Kumar, and A. Lakshmanarao, “Innovative way of identifying skin cancer model design with fcnn and lstm,” in 2024 IEEE International Conference on Computing, Power and Communication Technologies (IC2PCT) , vol. 5, 2024, pp. 287– 291

    2024

  5. [12]

    Skin lesion segmentation based on multi-scale attention convolutional neural network,

    Y . Jiang, S. Cao, S. Tao, and H. Zhang, “Skin lesion segmentation based on multi-scale attention convolutional neural network,” IEEE Access , vol. 8, pp. 122 811–122 825, 2020

    2020

  6. [13]

    Transfer learning and fine-tuned transfer learning methods’ effectiveness analyse in the cnn-based deep learning models,

    C. Öztürk, M. Ta¸ syürek, and M. U. Türkdamar, “Transfer learning and fine-tuned transfer learning methods’ effectiveness analyse in the cnn-based deep learning models,” Concurrency and Computation: Practice and Experience , vol. 35, no. 4, p. e7542, 2023. [Online]. Available:...

    2023 doi

  7. [14]

    Run, don’t walk: Chasing higher flops for faster neural networks,

    J. Chen, S.-h. Kao, H. He, W. Zhuo, S. Wen, C.-H. Lee, and S.- H. G. Chan, “Run, don’t walk: Chasing higher flops for faster neural networks,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2023, pp. 12 021–12 031

    2023

  8. [15]

    FALCON: FLOP-aware combinatorial optimization for neural network pruning,

    X. Meng, W. Chen, R. Benbaki, and R. Mazumder, “FALCON: FLOP-aware combinatorial optimization for neural network pruning,” in Proceedings of The 27th International Conference on Artificial Intelligence and Statistics , ser. Proceedings of Machine Learning Research, S. Dasgupta...

    2024

  9. [16]

    Automated skin lesion classification using ensemble of deep neural networks in isic 2018: Skin lesion analysis towards melanoma detection challenge,

    M. A. A. Milton, “Automated skin lesion classification using ensemble of deep neural networks in isic 2018: Skin lesion analysis towards melanoma detection challenge,” 2019. [Online]. Available: https://arxiv.org/abs/1901.10802

    2018 arXiv

  10. [17]

    A comparative analysis of deep learning and hybrid models to diagnose multi-class skin cancer,

    I. N. Nawrin, T. T. Trina, R. Rahman, and A. A. Rasel, “A comparative analysis of deep learning and hybrid models to diagnose multi-class skin cancer,” in 2023 International Conference on Next-Generation Computing, IoT and Machine Learning (NCIM) , 2023, pp. 1–6

    2023

  11. [18]

    Melanoma segmentation: A framework of improved densenet77 and unet convolutional neural network,

    M. Nawaz, T. Nazir, M. Masood, F. Ali, M. A. Khan, U. Tariq, N. Sahar, and R. Damaševi ˇcius, “Melanoma segmentation: A framework of improved densenet77 and unet convolutional neural network,” International Journal of Imaging Systems and Technology , vol. 32, no. 6, pp. 2137–2...

    2022 doi

  12. [19]

    Using imagenet xception model to identify skin cancer and non-skin cancer image classification,

    M. I. H. Abir, A. Hossain, and T. Salam, “Using imagenet xception model to identify skin cancer and non-skin cancer image classification,” in 2024 IEEE International Conference on Computing, Applications and Systems (COMPAS), 2024, pp. 1–6

    2024

  13. [20]

    Skin cancer classifi- cation and detection using vgg-19 and desnet,

    A. Barbadekar, V . Ashtekar, and A. Chaudhari, “Skin cancer classifi- cation and detection using vgg-19 and desnet,” in 2023 International Conference on Computational Intelligence, Networks and Security (ICCINS), 2023, pp. 1–6

    2023

  14. [21]

    Benign and malignant skin lesion detection from melanoma skin cancer images,

    S. Sharma, K. Guleria, S. Kumar, and S. Tiwari, “Benign and malignant skin lesion detection from melanoma skin cancer images,” in 2023 International Conference for Advancement in Technology (ICONAT) , 2023, pp. 1–6

    2023

  15. [22]

    Diagnosis of skin cancer via transfer learning with combined channel attention and spatial atten- tion,

    Z. Maqbool, S. Pumrin, and N. Panitantum, “Diagnosis of skin cancer via transfer learning with combined channel attention and spatial atten- tion,” in 2024 28th International Computer Science and Engineering Conference (ICSEC), 2024, pp. 1–5

    2024

  16. [23]

    Deepskin: A deep learning approach for skin cancer classification,

    H. L. Gururaj, N. Manju, A. Nagarjun, V . N. M. Aradhya, and F. Flammini, “Deepskin: A deep learning approach for skin cancer classification,” IEEE Access, vol. 11, pp. 50 205–50 214, 2023

    2023

  17. [24]

    Dermatologist-level classification of skin cancer using cascaded en- sembling of convolutional neural network and handcrafted features based deep neural network,

    A. K. Sharma, S. Tiwari, G. Aggarwal, N. Goenka, A. Kumar, P. Chakrabarti, T. Chakrabarti, R. Gono, Z. Leonowicz, and M. Jasi´nski, “Dermatologist-level classification of skin cancer using cascaded en- sembling of convolutional neural network and handcrafted features based dee...

    2022

  18. [25]

    Enhanced magneto-optic imaging based on lstm-cnn model and multi-modal fusion,

    S. Huang, G. Jin, and M. Jiyun, “Enhanced magneto-optic imaging based on lstm-cnn model and multi-modal fusion,” in 2021 IEEE International Conference on Emergency Science and Information Tech- nology (ICESIT), 2021, pp. 612–621

    2021

  19. [26]

    Cu-net: A new improved multi-input color u- net model for skin lesion semantic segmentation,

    R. Ramadan and S. Aly, “Cu-net: A new improved multi-input color u- net model for skin lesion semantic segmentation,”IEEE Access, vol. 10, pp. 15 539–15 564, 2022

    2022

  20. [27]

    Analysis and prediction of energy consumption in neural networks based on machine learning,

    X. Qi, T. He et al. , “Analysis and prediction of energy consumption in neural networks based on machine learning,” Academic Journal of Computing & Information Science , vol. 7, no. 4, pp. 90–97, 2024

    2024

  21. [28]

    Mcmc: Multi- constrained model compression via one-stage envelope reinforcement learning,

    S. Li, J. Chen, S. Liu, C. Zhu, G. Tian, and Y . Liu, “Mcmc: Multi- constrained model compression via one-stage envelope reinforcement learning,” IEEE Transactions on Neural Networks and Learning Sys- tems, 2024

    2024

  22. [29]

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

    P. Tschandl, “The HAM10000 dataset, a large collection of multi- source dermatoscopic images of common pigmented skin lesions,”

  23. [2018]

    Available: https://doi.org/10.7910/DVN/DBW86T

    [Online]. Available: https://doi.org/10.7910/DVN/DBW86T

    doi

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.