MPox-DenseConvNet: A Transfer Learning Based Convolutional Neural Network for Monkeypox Detection and Assessment using Color Models

Abstract

Monkeypox, a zoonotic orthopoxvirus, unintentionally produces smallpox-like sickness in people, though with a far lower death rate. Although Deep Networks have been extensively used for visual inspection of such diseases, the majority of works have frequently relied their analysis on the results produced by a particular network without taking the responsibility of the color channels to classify findings into account. Deep learning has recently been shown to have enormous potential for image-based diagnosis, including the detection of skin cancer, the identification of tumor cells, and the COVID-19 patient diagnosis through chest radiography. As a result, a similar application may be used to identify the sickness associated with monkeypox as it impacted human skin. This image can then be obtained and employed to identify the illness. This work focused on investing the prominent color channel for Convolution Neural Network (ConvNet) based monkeypox classification using skin images. For this purpose, a transfer learning-based classification architecture named MPox-DenseConvNet with fine-tuning is designed. Three color channels namely RGB, HSV, and YCbCr are analyzed using the proposed MPox-DenseConvNet. The outcomes demonstrated that the color channel employed had an impact on the performance of the classification. The results also confirmed that the HSV color channel has outperformed all the color channels taken into consideration.