mmm1712/edge-ai-object-tracking-camera

This is a DIY hardware project for a small camera that physically pans and tilts to follow a moving object, entirely on its own without needing a laptop or any other computer nearby. The creator's first version streamed video from a camera module to a laptop, where AI software running on the laptop figured out where the object was and sent movement commands over to a separate microcontroller. That worked, but the camera was useless without the laptop connected. In this upgraded version, the object detection itself runs directly on a small AI vision module called the Grove Vision AI V2, so no laptop or external computer is needed at all. That module detects where the object is in the camera's view and sends the position to a XIAO ESP32-S3 microcontroller, which compares that position to the center of the frame and moves two small servo motors to turn the camera toward the object. The further off center the object is, the faster the camera turns, and small built-in limits keep the movement smooth rather than jerky. Building it requires assembling specific hardware: the Grove Vision AI V2 kit, a XIAO ESP32-S3 board, a camera module, two micro servos, a servo driver board, an external power supply for the servos, and a custom 3D-printed pan and tilt body whose design files are included in the project. The firmware is written for the Arduino IDE using two specific libraries, and setup involves wiring the parts together, uploading the provided firmware sketch, and adjusting tracking settings like servo limits and movement sensitivity near the top of the code. The author describes this as a working prototype rather than a finished product, with a list of planned improvements including cleaner wiring, battery power, and a sturdier enclosure. The project is released under the MIT License.