Low-Power Edge AI Inference BoardVAB1000
Key Features
- Low Power Consumption
- 3W (Equipped with a Japan-made AI inference ASIC.)
- Compact Design
- Board size: 45 mm × 65 mm
- Programmable Platform
- Runs Ubuntu OS and supports custom AI models and application development.
- High-Speed Image Processing
- Hardware-accelerated image processing with OpenCV support.
- Optional Expansion Kit
- We also offer an optional camera kit with the AI board pre-installed for easy deployment.
Application
A compact edge AI solution for adaptive on-site processing
Hardware Specifications
| Items | Specifications | Comments | |
|---|---|---|---|
| AI Engine | 2K Low-Power ASIC | Compact and battery-operable design. Made in Japan | |
| External Interface | POWER1 | USB Type-C bus power input | Power supply only (no data transmission) |
| POWER2 | 5V DC power input | For use in combination with POWER1 | |
| USB HOST | USB Type-C USB host port Wireless LAN adaptor | Supports video input from USB cameras | |
| UART | USB Micro-B UART interface for the onboard CPU in AI LSI | ||
| Audio interface | Input: MIC, AUX Output: Headphone |
Dedicated analog audio interface (mono) | |
| Sensor interface | CMOS image sensor interface | ||
| JTAG | JTAG interface for the AI LSI | ||
| Storage | For maintenance purposes | ||
| Interface extension *1 | Video monitoring via external expansion board | 5.0 V power, video output signals, and GPIO signals | |
| Ethernet interface supported via external expansion board | 1.8 V / 3.3 V power and Ethernet PHY interface | ||
| RTC Battery | RTC backup battery connector | Real-time clock (RTC) | |
| Operating Conditions | Temperature | -20℃ to +65℃ (ambient) | |
| Humidity | 30 to 90％ RH | Non-condensing | |
| Power Consumption | 5.2 W max. | ||
| Dimensions | Board Size | 65 mm x 45 mm | |
- *1
- The interface expansion function allows additional interfaces to be implemented using a separate expansion board.
Required circuits (such as PHYs) and connectors must be implemented on the expansion board.
Software Specifications
| Items | Specifications | Comments |
|---|---|---|
| OS | Ubuntu 22.04 LTS | |
| Library | OpenCV 4.5.1/GStreamer 1.20.1 | When using the OpenCV acceleration API, please use OpenCV 4.5.1 instead of the default OpenCV 4.5.4 package included with Ubuntu 22.04 LTS. |
| API | OpenCV Hardware Acceleration API | Additional APIs will be supported in future releases. |
Supported AI Models
| Category | Models |
|---|---|
| Object Detection | YOLOv3, YOLOX-Tiny |
| Pose Estimation | Lightweight Human Pose Estimation |
| Age Estimation | PINT00309 (Age/Gender Recognition) |
| Gender Estimation | |
| Gaze Estimation | Gaze Estimation |
Block Diagram
- *2
- Reduced Gigabit Media-Independent Interface
- Additional built-in AI models will be added and updated in future releases.
- Supported OpenCV functions are available upon request.
- Recommended LTE/Wi-Fi/Bluetooth communication devices are available upon request.
- Recommended CMOS image sensors compatible with the sensor interface are available upon request.
- The interface expansion function enables additional interfaces through a separate expansion board. Required circuits (e.g., PHYs) and connectors must be implemented on the expansion board.
- This content is subject to change without notice.