Real Time Application of Deep Learning Approach in Exogenous Skin Problem Identification

摘要

The proposed research work provides an easy way to the dermatologists and other medical professionals to acknowledge any given skin problem in a better way. The proposed work based on deep learning algorithm aims to make up the need of dermatologists in India. In the landmass sized country roughly 11,000 derma experts are employed, which means less than one derma experts for 100,000 patients. This work aims to help relieve the circumstances. It can diagnose five types of skin problems like redness. swelling. puss. cold sore. and tinea versicolor. The proposed system assure the patients with an option to self-check their signs or symptoms like redness swelling etc that makes this work a real time application. The patient can also then associate to a dermatologist online for consultation within a momentary span of time by unloading their photos to hospital website, and dermatologists give guidance by analyzing the photos with faster R-CNN deep learning algorithm. The system uses tensorflow object detection framework with faster RCNN configuration for skin problem like redness, pus and swelling assessment and successfully estimates these skin problem depending on maximum scores from the trained model. The main aim of this system is to achieve maximum accuracy of dermatological problem prediction.