Using vision as a sensor to monitor the state of legacy, non-connected devices and integrating them into my smart home.
I've been dabbling in home automation for a few years, using the fairly common MING stack (MQTT, influxdB, node-red, Grafana), I've also built a few custom ESP based sensors.
I'm now exploring using computer vision as a sensor to monitor things that aren't connected or 'smart' enabled yet.
I've trained an object detection model to watch my CCTV IP cameras, it finds the locks on my back door and uses a second model to classify the state. I then do the usual publish the results over mqtt... the rest is history after that, getting fed into the MING stack.
Chapters:
00:00 Intro
00:11 Recap
00:35 Node-RED
02:02 Shelly IoT bridge
02:35 Node-RED Script
04:40 Google Integration by proxy
05:16 The Result
