Deep Learning-Based Classification of Cutaneous Viral Diseases: An Efficient Approach

Cutaneous viral diseases such as chickenpox and monkeypox are a significant public health concern, and accurate diagnosis and classification are crucial for effective treatment. On-time diagnosis, can not only lead to better treatment but can also help to limit the spread of the disease. With the same concern, this study emphasizes the image-based classification of three such diseases, namely, monkeypox, measles, and chicken pox, from the normal skin. For an effective classification of cutaneous viral diseases, CNN models have been employed. The suggested method entails initial image pre-processing, followed by the training of a convolutional neural network (CNN) using an extensive dataset of skin lesion images. The model's effectiveness is subsequently assessed on an isolated testing dataset. The methodology incorporates the implementation and comparative evaluation of two renowned CNN architectures, namely, VGG 16 and Inception V3, resulting in peak accuracies of 90% and 96%, respectively. Additionally, the fully connected layers are examined with four widely used activation functions: Linear, Tanh, Sigmoid, and ReLU.