If you are searching for a smartwatch that can easily be customized and is completely open-source then you may be interested in the new Bangle.js hackable smartwatch.
- IP68 Waterproof: up to 10m underwater
- Nordic 64MHz nRF52832 ARM Cortex-M4 processor with Bluetooth LE
- 64kB RAM 512kB on-chip flash, 4MB external flash
- 1.3 inch 240×240 16 bit LCD display with 2 zone touch
- GPS/Glonass receiver (UBlox)
- Heart rate monitor
- 3 Axis Accelerometer (with Pedometer and Tap detect)
- 3 Axis Magnetometer
- Piezo speaker and Vibration motor
- 350mAh battery, 1 week standby time
- 5 x 5 x 1.7 cm case, plastic with stainless steel ring
- Can be disassembled with just 4 screws
- 40kB RAM for program memory/variables
- Bluetooth 4.2 Advertising, Central and Peripheral mode support with built-in Nordic UART service
- Graphics library with Vector fonts, bitmap rotate & scale
- Tensorflow Lite for Microcontrollers AI
- Wear-leveled flash filesystem
- Heatshrink compression
- Upload functions written in C or ARM Assembler
- Built-in wireless debugging
- VT100 Terminal support on LCD
- Program with Web-based Web Bluetooth IDE or Node.js-based command-line tools
Developing software for embedded devices can be a real pain. You normally need wires attached to the device you’re developing for, and you need to install and set up a complex toolchain. Anything more than a few lines of code can take a minute or more to compile and upload to the chip, and this applies to every change you make. Many tools don’t even allow for debugging.
Tensorflow Lite AI
Tensorflow Lite for Microcontrollers is built into Bangle.js’s firmware so you can run machine learning models wirelessly on your wrist! This allows you to detect gestures using the accelerometer. If you’re feeling adventurous you can customize the data you collect and create or update machine learning models without even having to reboot your watch!
Simply train and test your model on your PC with Tensorflow, then create a tflite model and upload it to your watch.