Mobile Skin Disease Classification using MobileNetV2 and NASNetMobile

The people of Indonesia often suffer from three diseases: tinea versicolor, ringworm, and scabies. Most Indonesians cannot distinguish the type skin disease they because some have same characteristics patterns. Therefore, this study built an M-Health application to predict diseases. prediction process uses a deep learning model deployed in smartphone. challenge is limited number datasets data used personal requires permission patient or hospital. transfer overcome these limitations. two pre-trained models, MobileNetV2 NASNetMobile. To obtain with high accuracy, performed modifications architecture test results showed that performs best using rate 0.0005 activation function ELU. While NASNetMobile produces performance 0.0001 ReLU6. gallery smartphones show has accuracy 91.6%, while 88.9%. They were testing camera real-time, which resulted was not as accurate if gallery. Accuracy smartphone shows 75% when 72.2% MobileNetV2. will increase flashlight capturing objects. Based on flashlight, value increased 80.5%, changed 77.8%.