About AI Applications

This page explains the overview of AI Applications.

What is AI Applications?

AI Applications are sets of application source code, pre-build application binary and pre-trained AI model objects, which allow users to run AI Applications easily and quickly.
Users can select the category of applications and access the applications provided on GitHub.
Select category of AI Applications >

File Configuration

Application directory has the following structure.

Directory	Details
exe_*	Execution environment required when running application on each supported board.
src_*	Application Source Code (C++) for each supported board.
Other directory	Optional: Extra files that may be required other than above.
Makefile	Optional: Configuration to build the application. It may be included in the src_* directory.
README.md	Optional: Documentation of the application.

Prerequisites

To run AI Applications, users are required to have followings.

• Basic Linux terminal, Linux command and filesystem hierarchy knowledge.
  
• Git knowledge and skills of followings.
  • Clone the repository.
  • Checkout the branch.
  • Get access to files from a certain commit ID/tag.

Note AI Applications and AI SDK is based on CUI (Linux terminal/command).
For users who would like to use GUI instead of CUI, please use AI Navigator (Renesas e² studio plugin).
AI Navigator (e² studio plugin) >

Git Tag Rule

Necessary files for AI Applications are provided on GitHub repository.
Users can download them using git command on Linux PC.

In GitHub repository, application source code and related files are provided under the following rules.

Branches

• main
  • The latest application source code, binary files and documentations.

Tags

• vX.XX
  • Application source code, binary files and documentations of each version.
  • The latest version should be the same as the main branch.
  • E.g., To download the files of AI Applications v1.00, users can use following command.
    

    git clone -b v1.00 <GitHub repository url>

Note The link in AI Applications page guides you to the "main" branch of GitHub repository, which is the latest information.
Please be aware that latest AI Application requires latest RZ/V AI SDK.

Directory structure of each application in GitHub repository is explained in the File Configuration.

Camera input information

MIPI camera

RZ/V2L EVK

• Google Coral Camera is included in the Evaluation Board Kit.
  
  

RZ/V2H EVK

• Use the e-CAM22_CURZH camera, which supports FHD (1920x1080) resolution, provided by e-con Systems.
  
  • To use e-CAM22_CURZH camera in AI Applications, users are required to change the source code.
    Please contact e-con Systems for more details.
    
    
  • To use camera input resolution from VGA (640x480) to FHD (1920x1080), users are required to apply followng changes.
    1. Change camera input size in define.h and other files.
       
       
       Example:
       In R01_object_detection, change the define.h as shown below.
       Before

       #define CAM_IMAGE_WIDTH             (640)
       #define CAM_IMAGE_HEIGHT            (480)

       After

       #define CAM_IMAGE_WIDTH             (1920)
       #define CAM_IMAGE_HEIGHT            (1080)

       
       
    2. Change the AI model object directory name in define.h.
       AI model object for FHD input size is provided in the *_fhd directory.
       
       
       Example:
       In R01_object_detection, change the define.h as shown below.
       Before

       const static std::string model_dir = "yolov3_onnx";

       After

       const static std::string model_dir = "yolov3_onnx_fhd";

       Note 1 Some applications may require users to download the *_deploy.so (one of AI Model object files) from the Release page due to its filesize.
       Download the corresponding file into exe_v2h according to each README.md and rename it to deploy.so.
       Note 2 If the application does not provide the AI model object for FHD input size (*_fhd directory), you can use the same AI model object as for VGA input size.
       
       
    3. Other camera input size information when applying e-CAM22_CURZH camera change.

USB camera

For both RZ/V2L EVK and RZ/V2H EVK, AI Applications support USB camera with VGA (640x480) resolution in 'YUYV' (YUYV 4:2:2) format.

Licenses

AI models

See README.md provided in each AI Applications.

AI Applications Image

All images listed in AI Applications page are proprietary files owned exclusively by Renesas Electronics Corporation.

Reference

To select the AI Application, choose the category from AI Applications section of the top page.
Go to Home to select category >

Getting Started explains the whole procedure to build/run the AI Applications.
To build the application, see Step 6 of Getting Started.
Go to Getting Started Step 6 >

To run the application, see Step 7~8 of Getting Started.
Go to Getting Started Step 7 >

Back to Top >