Prokyber AI-on-the-Edge-Cam

Overview

The Prokyber AI-on-the-Edge-Cam is a purpose-built open-hardware board designed to run the AI-on-the-edge-device firmware, which uses TensorFlow Lite to digitize analog utility meters for water, gas, and electricity. Built around the ESP32-S3 N16R8 with 16MB flash and 8MB PSRAM, it provides enough memory and processing power for on-device neural network inference directly at the edge, eliminating the need for cloud-based image processing.

The board integrates Power over Ethernet via a WIZnet W5500 controller and Silicon Labs SI3404 PoE PD controller, enabling single-cable deployment where the same Ethernet cable provides both network connectivity and power. Four WS2812B addressable LEDs provide configurable camera illumination for consistent image capture in any lighting condition. A microSD card slot stores configuration, captured images, and firmware updates, while a Stemma QT (Qwiic-compatible) I2C connector allows easy expansion with external sensors like temperature or humidity modules.

With an ultra-low sleep current under 25 microamps and active current of only 200mA, the board is optimized for battery-powered deployments as well. Two 18650 cells can power daily meter readings for approximately one year. An optional LoRa shuttle shield adds LoRaWAN connectivity for transmitting readings in locations without Ethernet infrastructure. The hardware design is fully open-source under the CERN-OHL-P 2.0 license with schematics published on GitHub.