The Dog Eye Guardian App: From Image to Diagnosis with AI Insights

This research aims to present a specialized U-Net model that we designed for the segmentation of eye diseases in dogs so that it can be used in the preliminary assessment of diseases in veterinary ophthalmology. To train the U-Net model, a custom dataset was prepared for four common eye disease symptoms using veterinary expertise to label the images. The learned model was integrated into a prototype web application. The application is containerized using Docker for improved portability and scalability, ensuring it can run efficiently across platforms. The results showed the ability of the U-Net model to cost-effectively segment eye symptoms, and the implementation of the application enabled the application of the model to users and the collection of valuable additional data that is important for retraining the model. User experience, and their guidance and information make a valuable contribution in determining the direction of development of an application for diagnosing and treating eye diseases in dogs.