Unet based Image Segmentation of Sweat Droplets for Assisting Assessment of Atopic Dermatitis

In this paper, we describe an automatic analysis that identifies skin structure and sweat droplets from skin surface images. The analysis uses a discriminator based on deep learning and has succeeded in producing results that in some cases exceed the diagnosis of a medical experts. The discriminator is built based on U-Net, which is used for image segmentation.