Detection of Human Skin Burn using ABNet Model

Human skin burns are severe injuries that require precise diagnosis for efficiently treating and managing burnt skin. The process of diagnosing burns with the traditionally followed method is a time-consuming task and requires professional training. On the other hand, machine learning techniques promoting smart healthcare are a boon for healthcare. In this research, an innovative ABNet model is developed that consists of three primary layers i.e. Conv2D layer, Max-pooling2D layer, and Relu layer along with SE Block after each set of three layers to determine the degree of the burn and need for grafting. The proposed model showed remarkable results of 98.25% accuracy for binary-class classification and 96.08% accuracy for multiclass classification. ABNet model can be used on cloud servers to quickly and more accurately screen burn patients ensuring the healthy lives of affected ones.