MonkeyPox, Measles and ChickenPox Detection through Image-Processing using Residual Neural Network (ResNet)
Kapil Sharma, Kishlay, Vinod Kumar, Mayank Mittal
Abstract
Timely detection of Measles, Mpox (Monkeypox), and Chickenpox can help in effective treatment and computer-aided decision-making for faster diagnosis. In today’s scenario, measles is spreading throughout India and Mpox is spreading worldwide. This issue requires a solution because it has the potential to grow into a major problem in the future and even cause a pandemic. The contribution of Deep Learning to large-scale medical data research has been invaluable, offering new approaches and opportunities for diagnostic techniques. In this research-based project, we tried to explore how deep learning models such as CNNs and Residual Neural Networks (ResNets) using datasets containing images of skin lesions can be used to classify various Measles, ChickenPox and MonkeyPox. We tried to highlight the challenges (including dataset size and quality) in using the publicly available “Monkeypox Skin Lesion Dataset (MSLD)” to develop our Proposed ResNet model and then compare the results of VGG 16, Inception V3 and CapsuleNet. Images used in the dataset are collected through various sources by web scraping. To increase the sample size, we have used data augmentation. We have developed a custom RESNET-18-based model and compared it with various other models such as VGG-16, RESNET-50, InceptionV3, Ensemble etc. Our model achieved an accuracy of 84.59% in the classification of monkeypox, measles and chickenpox.