Scopus Indexed Publications

Paper Details


Title
Deep Learning-Based Approach for Detecting Multiple Skin Diseases: A Comprehensive Study

Author
Aklema Shorna, Abida Sultana Ave, Md Abrar Hamim, Nourin Jahan Sowrna, Sabrina Bari Shital,

Email

Abstract

Presents a deep learning-based approach for classifying five types of skin diseases: Vitiligo, Rosacea, Impetigo, Ringworm, and Hives. A transfer learning method was employed using MobileNetV3 (Small & Large), InceptionV3, Xception, DenseNet201, and DenseNet121. To enhance model performance, the dataset underwent preprocessing (image resizing, normalization) and augmentation (rotation, flipping, zooming, and brightness adjustment). Experimental results show that DenseNet121 achieved the highest accuracy of 94.40% with a test loss of 0.63, demonstrating its effectiveness in skin disease classification. This study highlights the potential of transfer learning for automated dermatological diagnosis and suggests future improvements, including expanding datasets with diverse skin tones and mobile-based implementations.


Keywords

Journal or Conference Name
2025 1st International Conference on Computational Intelligence Approaches and Applications, ICCIAA 2025

Publication Year
2025

Indexing
scopus