The defect detection system uses a neural network to monitor your printer's webcam stream. This state-of-the art computer vision model was developed at Google for use on mobile devices like the Pixel phone.
A compression technique called quantization reduces the footprint of the model, enabling real-time performance directly on your Raspberry Pi.
This system was engineered with offline reliability and limited bandwidth in mind. A 3D printer is a hazardous home appliance, like a gas oven.
You wouldn't trust a "smart" internet-connected stove, right? (please don't, see below for examples of what could happen 😬)
Note: the current beta version requires an internet connection for analytics & telemetry. An offline mode will be available in later releases.