From a Weekend Hack, to a Future-Proof IoT Device
summary
A weekend Arduino prototype with a four‑channel gas sensor becomes a portable TinyML device: data captured at 10 Hz sampling, features extracted (min/max/mean/std) and trained with Edge Impulse, running inference with TensorFlow Lite Micro in a few milliseconds on a 200 MHz Cortex‑M4. Code and 3D enclosure live on GitHub; migration to Zephyr adds threads, zbus messaging, LVGL GUI, simulator support and MQTT — demonstrated on SAMD51 and M5Stack ESP32 and used for restroom‑monitoring IoT scenarios.
AI-written summary · about
chapters (13)
- 0:00Artificial nose project intro
- 1:24Pandemic baking inspiration
- 3:35ML to sensor data
- 6:47Public reaction and imposter syndrome
- 9:17Real impact on patient
- 13:12Arduino development challenges
- 15:38Migrating to Zephyr
- 18:49App with Zephyr and zbus
- 21:57Live demo artificial nose
- 24:09IoT air quality use cases
- 26:58Digital twins in facilities
- 28:44Future-proofing with Zephyr
- 31:00Q&A
related talks
39mFrom a Weekend Hack, to a Future-Proof IoT Device
35mData science and Internet of Things on the edge
16mEdge Computing and next-gen of IoT Sensors
18mManagement of IoT TinyML Devices
54mSmarter Everything with ML & the IoT
01h 01mML and the IoT: Living on the Edge
45mUsing Edge Machine Learning to save power and development time
59mML and the IoT: Living on the Edge