Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{203145,
        author = {Nayan Sahebrao Jadhav and S.H.Jadhav and Dr. Syed Sumera Ali and Dr. D.L. Bhuyar and Dr.G.B.Dongre},
        title = {Deep Learning-based Multi-Class Skin Disease Detection using CNN and Transfer Learning},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {12},
        pages = {11730-11737},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=203145},
        abstract = {The rising cases of the skin diseases and the issues of diagnostic complexity of the similar lesion types that look similar, has prompted the design of automated computer-aided diagnostic systems to assist in clinical decision-making. In this work, the authors suggested a deep learning-based multi-class skin disease classification system that combines convolutional neural networks with transfer learning to increase the accuracy of the diagnosis and the applicability of the proposed predictions in a realistic data environment. The publicly available benchmark dataset was pre-processed into dermoscopic images, with normalization and abundant data augmentation being used to reduce the imbalance among the classes and enhance the robustness. Fine-tuning of a pre-trained CNN backbone with a custom classification head allowed fine-tuning of discriminative feature learning across various clinically relevant skin disease classes. The suggested framework was tested with the help of overall performance measures, such as accuracy, precision, recall, F1-score, macro-averaged AUC, analysis of performance on a class-by-class basis, as well as evaluating the error pattern through the use of a confusion matrix. The overall classification performance of the experiment was high, and the separability of the classes between the disease types was uniform with significant misclassification patterns between the visually similar lesions, especially between early-stage melanoma and benign nevi. Ablation analysis also supported the fact that transfer learning and data augmentation were vital performance and generalization factors, which performed much better than models trained on baseline. These results revealed that the presented framework can provide a solid and scalable framework to detect multiple-class skin disease and has the potential to become a computer aid diagnostic tool to aid the initial dermatological screening, whereas subsequent studies should consider the inclusion of lesion-centric attention mechanisms and multimodal clinical data to enhance the reliability of high-risk malignant cases.},
        keywords = {Deep learning, convolutional neural networks, transfer learning, skin disease classification, dermoscopic image analysis, computer-aided diagnosis.},
        month = {May},
        }