Overview

Description

Updated: Jan. 15, 2023
Renesas Electronics Corporation

It provides e-AI (embedded artificial intelligence) that infers and executes many of the open source AI frameworks at endpoints using RZ/G2.

Customers can use a collection of Yocto metalayers that allow native inference execution on the RZ/G Series reference platform.

Customer can also quickly evaluate the RZ/G2 series using an image-based sample benchmarking application for product recognition in shopping basket.

Product Overview

The RZ/G series AI BSP incorporates the well-known FOSS (Free and Open Source Software) AI frameworks ArmNN, ONNX Runtime, and TensorFlow Lite into VLP. Customer can perform inference using the results learned on the RZ/G2 platform.

AI BSP Package also includes a sample benchmarking application for each framework that tests the execution performance of some well-known pre-trained image classification models.

Customers can immediately use Linux with AI inference on RZ/G2 boards.

Customer can also try a demo for evaluating RZ/G2 products. Customer can immediately evaluate your RZ/G2 product in the demo.

Features

  • Providing a collection of Yocto metalayers (meta-renesas-ai)
  • For support frameworks are ArmNN, ONNX, Runtime and TensorFlow Lite
  • Easily create an RZ/G series environment by building with VLP or BSP. VLP

Functions

  • Yocto metalayers (meta-renesas-ai)
  • Support framework: ArmNN(v22.02), ONNX Runtime(v1.8.0), TensorFlow Lite(v2.5.3)
  • Sample benchmark application for evaluation

Target Reference Boards

  • RZ/G2 Group
    • HopeRun Technology HiHope RZ/G2H platform (hihope-rzg2h) Rev 4.0
    • HopeRun Technology HiHope RZ/G2M platform (hihope-rzg2m) Rev 4.0 (*1)
    • HopeRun Technology HiHope RZ/G2N platform (hihope-rzg2n) Rev 4.0
    • Silicon Linux RZ/G2E evaluation kit (ek874) Rev C
    • Silicon Linux RZ/G2E evaluation kit (ek874) Rev E
    • RZ/G2L Evaluation Board Kit PMIC version (*2):
      • RZG2L SMARC Module Board v02
      • RZ SMARC Series Carrier Board v04
    • RZ/G2LC Evaluation Board Kit
  • (*1) There are 2 versions of RZ/G2M LSI devices. There are version 3.0 and 1.3 (hereinafter referred to as “RZ/G2M v3.0” and “RZ/G2M v1.3”).
    (*2) “RZ/G2L Evaluation Board Kit” and “RZG2L Evaluation Board Kit PMIC version” includes the RZG2L SMARC Module Board and the RZ SMARC Series Carrier Board.

History

Dec. 29, 2021: New release
Jan. 15, 2023: v5.0.0

Download

Please check the Release Notes below and download the required files.

If you don’t need detailed explanation, please use GitHub.

Support Policy

As a general rule, Renesas does not provide support for this Package.

Blog

Embedded Artificial Intelligence Inference with Renesas RZ/G2

Target Devices

Design & Development

Related Boards & Kits

Boards & Kits

Videos & Training

Introduction of RZ/G AI BSP

This video introduces the AI BSP of the RZ/G series.

Edge-AI can be easily realized with Renesas RZ/G MPU by using OSS-based AI BSP.

00:00 Opening
00:10 The Features of RZ/G AI BSP
00:40 RZ/G2 AI Edge AI DEMO
01:00 Pose Estimation
01:50 Face Detection and IRIS Detection
03:10 Closing 

Related resources: