Azure IoT Edge + Movidius Neural Compute Stick ML Demo App

This app has 3 IoT Edge modules:

• Image Server - this module uses the Linux GStreamer API to interface with a Video4Linux compatible camera to sample image frames and make them available on a WebSocket endpoint using JPEG encoding. The WebSocket server listens on port 3012 by default. This has been written using the Rust programming language.

• Machine Learning Server - this module uses the Movidius Neural Compute Stick Python (NCS) SDK to interface with the NCS stick to run image pattern analysis jobs. This is also a WebSocket server written in Python and listens on port 8765 by default. This module has been adapted from the face matcher sample provided as part of the NCS SDK. In order to run this module you'll need to run through the setup instructions as described here. And you will also, of course, need the NCS USB stick itself.

• Web Frontend - this module is a static website that serves as a web interface for testing the system out. It uses nginx to serve the site and uses the JavaScript websocket API to interface with the image server and the ML server to sample image frames and evaluate them.