From a Weekend Hack, to a Future-Proof IoT Device
summary
An Arduino hack evolves into a Zephyr-based device that actually runs TinyML: a four‑channel gas sensor and a few‑kilobyte TensorFlow Lite model that recognizes coffee whiskey and lemon from minutes of training and noisy $20 sensors. Code and 3D case live on GitHub, sampling stays stable at 10 Hz sampling, inference runs in milliseconds on an ARM Cortex‑M4, GUI uses LVGL and zebus, threads keep acquisition, inference and MQTT uploads separate so the device reliably works across boards.
AI-written summary · about
chapters (12)
- 0:00Artificial nose project intro
- 3:12Arduino to tinyML and sensors
- 9:17Unexpected project applications
- 15:20Arduino challenges to Zephyr
- 21:46Live demo artificial nose
- 24:15IoT air quality use cases
- 27:30Zephyr advantages embedded
- 29:56Future possibilities open-source
- 31:00Q&A digital nose baking
- 32:47Q&A data collection models
- 35:44Q&A nose vs sensors
- 37:29Q&A next steps smell
related talks
39mFrom a Weekend Hack, to a Future-Proof IoT Device
14mKeynote: Leveraging Zephyr and ML to Bring Smart Devices to Market, Faster
54mSmarter Everything with ML & the IoT
31mKeynote: The next big thing in machine learning is tiny
01h 01mML and the IoT: Living on the Edge
42mPorting an AI Powered Wearable Health Monitor to Zephyr on Open Hardware
35mData science and Internet of Things on the edge
10mLightning Talk: Debugging Edge AI on Zephyr and Lessons Learned