Show HN: Pothole Detection System (YOLOv8 – FastAPI – Docker – React Native) (github.com)

A new project showcasing a YOLOv8-based pothole detection system blends advanced machine learning with user-friendly mobile technology. The application employs a fine-tuned YOLOv8 model, which is trained on a dedicated pothole dataset using Nebius Cloud's GPU resources. The inference system utilizes FastAPI within a Docker container, enabling efficient deployment and scalability. Users can interact with the system through a React Native mobile app, which captures road images and visualizes detected potholes in real-time with bounding boxes and confidence scores. This project is significant for the AI/ML community as it demonstrates a comprehensive machine learning lifecycle— from dataset preparation and model training to deployment and client integration. Key technical features include API key authentication for security, structured logging, and GPU/CPU auto-selection in the inference server. The project emphasizes the practical application of AI in infrastructure maintenance, potentially paving the way for more intelligent transportation solutions that can enhance road safety and reduce maintenance costs. With endpoints for image inference and health checks, this solution enhances the accessibility and utility of machine learning models for real-world applications.