Adaptive system and method for classification of skin abnormality images using an imbalanced dataset

Abstract

The present invention relates to an adaptive system for classification of skin abnormality images (100) using an imbalanced dataset, characterized in that, the system (100) comprising: a neural network module (110), a loss function module (120), and an optimizer module (130). The present invention further relates to a method for classification of skin abnormality images (200) comprising steps of: analyzing and classifying images of skin abnormalities (210) from an imbalanced dataset into multiple categories based on severity via a neural network module (110), improving classification performance (220) of the neural network module (110) on the imbalanced dataset by focusing on harder-to-classify cases, wherein the classification performance is improved via a loss function module (120) that utilizes a focal loss mechanism; and enhancing performance (230) of the loss function module (120) for better generalization on the imbalanced dataset, wherein the performance of the loss function module (120) is enhanced via an optimizer module (130) that implements AdaBelief optimizer. Figure 1