Overview
Description
The Trail Camera demonstration showcases the capabilities of the dynamically reconfigurable processor for AI (DRP-AI) subsystem in the RZ/V2L SoC along with a fast-booting Linux (<1.6s) implementation. It is a complex system that uses all major subsystems of the RZ/V2L. The on-chip Arm® Cortex®-M33 (CM33) MCU is used to capture images even before Linux has booted and the images are processed using the DRP-AI neural processing unit (NPU) to quickly identify objects in a complex scene. The ability to use a display to visually show the identified objects while the command line app shows a text-based log of inferences adds a lot of features that can each be extended in a product. The low power consumption is showcased by the simple use of a battery to power the entire system and keep the main board powered off till a motion is detected.
Features
- Fast boot Linux (<1.6 seconds)
- Object recognition using DRP-AI
- Heterogenous multi-processing in use: CM33 for image capture, ISP for preprocessing, DRP-AI for AI inference processing, and Arm® Mali for post processing rendering
- Inference results summarized on serial port
- Auto power on/off the main board
- Always on RA2E1 MCU tied to motion sensor application
- Low power consumption allows for battery-powered remote operations
Applications
Documentation
|
|
|
---|---|---|
Type | Title | Date |
Manual - Hardware | PDF 3.71 MB | |
1 item
|