What is spectacles-yolo-ml-example?

This project is an example Lens for Snap's Spectacles smart glasses that spots a real coffee cup through the wearer's camera and draws animated steam rising off it. It exists to show, step by step, how someone built an object detector for augmented reality glasses without doing traditional machine learning work themselves. The detection model behind it is trained through a public tool hosted on a service called Replicate. You open that page, fill in a short form, and press run, and the whole training pipeline happens automatically. It combines real photos captured through the Spectacles with computer generated training images made by an image generation model called FLUX, so you do not need to take hundreds of real photos yourself. Another AI model automatically draws the bounding boxes around the object in each training image, removing the need to label anything by hand. Underneath, an object detection model called YOLOv7 tiny is trained on this mixed data and exported into a small file format the glasses can run efficiently. Once trained, you swap in the resulting file to replace the example coffee cup detector, and update one line of code if you renamed what it detects. The Lens itself, built in Snap's Lens Studio tool, lets you place a virtual surface in front of you, look around until the camera spots the object, and see the steam effect appear once detection succeeds. It also includes a debug view that shows exactly what the model is seeing and where it thinks the object is, plus adjustable settings for how confident the model must be before it counts something as a detection, and how often it checks each camera frame to balance accuracy against battery life. The training pipeline, the code powering the detection, and the visual effect are all included as a working, swappable example rather than a general purpose product, and are credited to two named creators. The README does not state a license.