Maxlab's Tokay Lite Is an Open Source Espressif ESP32-Powered Camera Board for Edge AI
Designed around the ESP32-S3, this open source smart camera board includes night vision and on-device machine learning capabilities.
Canadian computer vision firm Maxlab has launched a crowdfunding campaign for the Tokay Lite, a smart camera development kit built around Espressif's ESP32 microcontroller — and boasting open source hardware designs, firmware, and software source code.
"Tokay Lite is an advanced AI [Artificial Intelligence] camera platform and ESP32-based development board with a wide range of applications," the company claims of its compact creation. "Whether you want to boost security, monitor wildlife, or keep an eye on your garden or crops, this versatile camera has you covered. Tokay Lite boasts an onboard MCU and sensor interface for edge AI processing. This enables real-time image analysis and decision-making right on the device, eliminating the need for an external computer."
The Tokay Lite device itself, brought to our attention by Linux Gizmos, is built around an Espressif ESP32-S3 microcontroller module — giving it a dual-core 32-bit Tensilica Xtensa LX7 processor running at up to 240MHz, 512kB of static RAM (SRAM) and 8MB of pseudo-static RAM (PSRAM), plus 16MB of off-chip flash, alongside Wi-Fi 4 and Bluetooth 5 Low Energy (BLE) radios.
To this, the company has added an Omnivision OV2640 image sensor — the same sensor on Espressif's own ESP-32-EYE camera module, which has only half the flash storage — with support for video at up to 15 frames per second and image capture at VGA, two megapixel, and three megapixel resolutions in RGB or JPEG formats. The camera includes an infrared filter cut system, allowing it to flick between filtered and unfiltered for night-vision applications — and IR LEDs are included on the board as standard, along with a passive infrared (PIR) sensor for motion detection and a light sensor for automatic nfight vision operation.
It's the device's on-device machine learning capabilities which Maxlab is looking to promote. The ESP32-S3 includes vector instruction extensions to boost the performance of computer vision and artificial intelligence workloads, supported by the 8MB PSRAM. Maxlab claims "effortless" connection to popular Internet of Things (IoT) and AI development platforms including Amazon Web Services (AWS) IoT, Edge Impulse, and ThingsBoard, along with support for TensorFlow Lite for Microcontrollers and Espressif's own ESP-DL deep-learning library.
"We’ve meticulously designed Tokay Lite with mass production in mind, optimizing it for seamless integration into the factory workflow," Maxlab claims of the future of the project's crowdfunding campaign. "These design enhancements significantly reduce production risks. With our streamlined approach, we’re confident there won’t be any production delays. In case of unexpected delays, we will be sure to inform everyone of any changes."
The Tokay Lite is currently funding on Crowd Supply at $89 plus shipping, with hardware expected to ship to backers in mid-May 2024. In the meantime the hardware design has been published to GitHub under the CERN Open Hardware License Version 2 — Permissive, while the software development kit has been published to a separate repository under the permissive Apache 2.0 license.