Building with TensorFlow Lite for microcontrollers
summary
Live demos show TensorFlow Lite for Microcontrollers running on an Arduino Nano 33 BLE Sense, receiving quantized 8‑bit models over WebBLE and executing gesture and sound classifiers that actually work — a drum kit and finger UI proved responsive. Training uses TensorFlow.js to collect IMU data, export a C model.h (example model sizes include the famous 13 kilobyte model), and flash preflashed kits; a $2,500 prize for five winners funds the challenge.
AI-written summary · about
chapters (11)
- 0:00Introduction to TinyML
- 4:50Understanding TinyML
- 14:06TinyML Overview Video
- 15:30The TinyML Process
- 20:18Setting Up the Environment
- 30:17Flashing the Arduino
- 35:15Motion Gesture Demo
- 37:12Finger Gesture UI
- 40:38Training Custom Models
- 46:47Real-World Applications
- 51:15Resources and Q&A
related talks
27mMachine Learning on the edge using TensorFlow Lite
27mWhat’s TinyML good for
54mSmarter Everything with ML & the IoT
32mAI for Mobile and IoT Devices: TensorFlow Lite
30mSimple JStures can go a long way
19mWhy Machine Learning on Microcontrollers?
55mEdgeAI - AI in the Palm of Your Hand
39mMachine Learning on Your Device: The Options