Vitiligo is one of the most common skin diseases in the world. According to the World Health Organization (WHO), the number of people suffering from vitiligo is growing year by year and vitiligo becomes a worldwide problem. In order to helping doctors with vitiligo diagnosis, we propose a vitiligo artificial intelligence diagnosis system. It is able to generate vitiligo images in Wood Lamp with high resolution and classify these images with high precision. In our system, we employ Cycle-Consistent Adversarial Networks (Cycle GAN) to generate images in Wood Lamp. What’s more, we use an advanced super resolution method, Attention-Aware DenseNet with Residual Deconvolution (ADRD), to improve the resolution of images. Finally, we obtain fantastic classification results with Resnet50. Our system is found to achieve the classification performance of 85.69% accuracy, which is increased by 9.32% compared with using Resnet50 to classify original images directly. The optimization and expansion of the system depend on the increase of data set and the improvement of system’s modules.

白癜风是世界上最常见的皮肤病之一。根据世界卫生组织（WHO）的报告，白癜风的患病人数正在逐年增加，白癜风已成为世界性的问题。为了帮助白癜风诊断医生，我们提出了一种白癜风人工智能诊断系统。它可以在高分辨率的木灯中生成白癜风图像，并可以对这些图像进行高精度分类。在我们的系统中，我们采用周期一致对抗网络（Cycle GAN）在Wood Lamp中生成图像。此外，我们使用高级超分辨率方法，带有残差反卷积的Attention-Aware DenseNet（ADRD），以提高图像的分辨率。最后，我们使用Resnet50获得了出色的分类结果。我们的系统被发现达到85的分类性能。精度为69％，与使用Resnet50直接分类原始图像相比，提高了9.32％。系统的优化和扩展取决于数据集的增加和系统模块的改进。