Computer Vision Based Skin Disorder Recognition Using Efficientnet

Abstract

Skin disorders have a vital impact on people's health and quality of life, it is essential to develop a reliable and accurate computer vision approach to recognize multiple skin disorders. In this paper, a new rapid recognition approach using Efficient Net has been introduced to diagnose twenty types of skin disorders. Initially, image augmentation techniques have employed, and then eight architectures of Efficient Net between B0 and B7 have trained using the transfer learning approach. To evaluate the performance of models, different experimental studies have employed using a test set of 6300 images of skin disorders dataset that makes the proposed approach more reliable and accurate. EfficientNet-B7 has achieved the highest accuracy 97.10% among all architectures but has taken longer training time. EfficientNet-B0 has taken the lowest training time and has achieved 93.35% accuracy. EfficientNet-B7 has also taken the lowest time in recognizing unseen new images of skin disorders accurately than others