JetCar: A miniature self-driving car based on Jetson Nano

JetCar is a miniature self-driving car based on Jetson Nano. It can navigate on street maps and follow stop text and directional arrows. JetCar is a 3D printed car based on NVIDIA Jetson Nano development board that requires very little additional electronics to drive. Using only camera stream as input, it can not only follow street markings but also automatically turn left or right at allowed intersections. Through machine learning, it can recognize directional arrows, stop text and stop lines on the street. The model architecture is U-Net. It creates very intuitive class images and is processed in the firmware, which is written in Python and controlled by Jupyter notebooks. The user connects to the car from the host computer via WiFi and the operator only needs to request a direction change for the next intersection. But it will only turn if the direction is not restricted by the directional arrows on the street. The project includes tools for mechanical design, electronic design, firmware and data preparation, model training, and street map generation. The documentation describes all parts in detail. All source code and binaries are available on GitHub.