Tokay PRO: Edge AI Camera
Pushing Forward Industrial High-performance Edge Computing
Tokay Pro – high performance modular camera
Over the last year, we have been working on bringing up the Tokay Pro advanced camera. This camera allows connecting up to 3 different sensors ( video, thermal, and spectral, for example) for the most ambitious projects. The camera can detect and track multiple objects in a single frame while running high-performance computer vision models.
Specs and datasheets
We gathered the answers to some popular questions below.
If you can’t find your question below feel free to contact us, and we’ll be happy to help.
What CPU are we using for the camera?
Tokay Pro’s CPU is NXP i.MX 8M Plus with on-board AI-accelerator.
What kind of camera sensor are you using?
Our camera supports multiple sensors, and we can attach up to 3 different sensors simultaneously on the device. What it means is our camera can support thermal, spectral, 3D, or visible spectrum sensors at the same time. The default option is IMX219, but it can be easily replaced with any other module, as long as it supports Raspberry Pi camera connector.
Is there a motion sensor on the device?
We are using PIR: AS312 motion sensor,
What kind of SOM are you using?
The system-on-a-module. Contains CPU, RAM and FLASH memory and nice form factor to plug that in into the carrier board. Comes with WiFi and BT capability. We use DART-MX8M-PLUS.
Do you have a microphone on the camera?
Yes! Our camera has a microphone. It is a MEMS mic to record audio.
Do you have a GPIO expander?
We can extend pins numbers using GPIO expander. If you feel you need additional pins, we have a chip for you.
Chip i.MX 8M Plus Block Diagram
Cameras are available for purchase, and we can ship the sample within a few weeks.
This year’s goal is to develop a high-performance devkit using one of NXP AI-accelerated chips based on Cortex-A53.
While Tokay Lite is optimized for ultra-low power usage, the next iteration of our devkits focuses on high performance.
We got feedback from folks at Hacker News and Hackaday , a lot of people were interested in the device that will be able to do multi-object recognition with 4k/60 fps video stream, and with the current AI accelerated video chip market being where it is right now, the NXP is the logical setup to use for this project.
The Spec We Aim For
With chips like i.MX 8M Plus, it is possible to achieve real-time detection of multiple objects (traffic counting, accident detection, object avoidance.) More importantly, this chip class is manufacturing-ready and can be used in real-life development without switching between devkit and production chips. If you want to develop something using a camera and an AI chip, give us a shout.
While devkits are awesome , we aim to show an actual demo project using edge computing, and we choose Inventures as our platform for May 30. We will showcase an interactive stand-alone hardware platform that helps physical businesses to remove friction in stores.
| Type | Characteristic |
|---|---|
| Onboard SoC | NXP |
| CPU Frequency | 800MHz |
| Camera Interface | MIPI-CSI |
| Night Vision | Yes |
| Stock Camera Sensor | IMX219 |
| Video Resolution | 8MP – 3840×2160 |
| Raw Video Frame Rate | 30 FPS |
| Video Encoding | H264 @ 30 FPS |
| Available RAM | 2Gb |
| Available Storage | 4Gb eMMC |
| External Storage | SD Card |
| Sensors | Light Sensor, Passive IR (Motion Detection) |
| Connectivity | WiFi, USB-C |
| Power Options | USB-C with backup battery |
