Recently I built for a group of IoT students here in Barcelona. The project combines the MING stack (MQTT, InfluxDB, Node-RED and Grafana) with Edge Impulse for computer vision.
We trained a ML model with Edge Impulse for Raspberry Pi running balena that could be automatically deployed, and used a camera to detect objects from the ML model in real-time.
Inference results were published to the MQTT broker (check cam/capture.py). Later in Node-RED we created a flow to subscribe to the MQTT topic where the inference results were being published, we stored them on influxDB and visualized the results on Grafana.
We also connected the inference results with a Telegram bot via Node-RED to use instant notifications in the mobile app.
Feel free to explore and share feedback 