DRP-AI Demos

From Renesas.info

RZV2 Reference Application Demos can be found in the links below. The links to the DRP and ISP Driver Package ZIP files include the demo code listed on this page. The Evaluation Software, Implementation Software, and Application Demos are included in the DRP-AI Package. The ISP demos are included in both the DRP-AI and and ISP package..

  • Renesas Implemented AI Models
    • YoloV2/V3. TInyYolov2/v3, Resnet, Moblenet, mmpose, and DeepLabv3
    • Refer to DRP-AI Translator page for other models.

DRP-AI Demo Binary

RZV2L DRP-AIDemoPre-builtImage.png
  • You can evaluate the AI demos such as Object Detection and Pose Estimation very easy after you get a RZ/V2L-EVKIT[1]

  • There is a pre-built Linux SD Card Image posted on renesas.com.
  • No Yocto building required.
  • No Linux PC required.

Pre-trained Free AI Application Libraries for RZ/V2L

Don't want to spend time training???

🆕 Renesas Offers Custom Pretrained models that can be used by customers for end products!

For more information go Pre-trained Free AI Application Libraries for RZ/V2L.

DRP-AI Sample Application Demos

The RZV2L, RZV2MA, and RZV2M DRP-AI Sample Application Demos are included in the DRP-AI Support Package.

Example for the RZV2M DRP-AI Support Package

    ├───rzv2m_drpai-sample-application   <- DRP-AI Demos
    ├───rzv_ai-evaluation-software       <- DRP-AI Evaluation Software
    ├───rzv_ai-implementation-guide      <= DRP-AI Implementation Guide
    │   ├───appendix
    │   ├───darknet_yolo
    │   ├───mmpose_hrnet
    │   ├───pytorch_deeplabv3
    │   ├───pytorch_mobilenet
    │   └───pytorch_resnet

DRP-AI Evaluation Software

The AI Evaluation Software is a precompiled application software that evaluates any DRP-AI inferences generated from the DRP-AI Translator.. The software loads and runs the DRP-AI inference in the DRP-AI hardware for evaluation. Runtime performance of the inference is provided. There is no post processing. Instead a raw binary is generated for post processing later.

NOTE: Source Code is not provided but the Application Software code that loads and runs the DRP-AI inference is the same. Please refer to the Application Software

  • RZ/V2M
    • rzv2m_ai-evalutation-software
  • RZ/V2L
    • rzv2l_ai-evalutation-software

DRP-AI Implementation Software

The Implementation Software are development tools that provide the following support.

  • PyTorch Scripts of Supported Renesas AI Models.
  • Python script example for converting PyTorch script to onnx
  • DRP-AI Translator configuration YAML files
  • Post Processing of the Evaluation Software for Supported Renesas AI Models.
  • Pre generated ONNX file and DRP-AI inference files.

Implementation Software can be found in the sub folder shown below of the RZ/V2L and RZ/V2M DRP-AI Support Package.

  • RZ/V2M
    • rzv2m_ai-implementaion-guide
  • RZ/V2L
    • rzv2l_ai-implementaion-guide

DRP-AI Application Software

The DRP AI Application Software are C language make file projects. These demos provide examples for running DRP-AI inference, Camera API ( USB and MIPI ), Weston Display API, H.264 network streaming, AI Preprocessing, and AI post processing. For detailed information refer to the documentation included in the package.

NOTE: Applications ending with _img do not use the camera or display. They read the image file, that is in the same directory as the demo executable then outputs the inference results to Terminal.

Application Software can be found in the sub folder shown below of the RZ/V2L and RZ/V2M DRP-AI Support Package.

  • RZ/V2M
    • rzv2m_drpai-sample-application
  • RZ/V2L
    • rzv2l_ai-sample-application
    • NOTE: The TinyYoloV2 demo "app_tinyyolov2_cam" does not use the RZV2L ISP.

DRP-AI and ISP Demos

This software is included in the DRP-AI Application Software. This software requires that the ISP has been installed. Below is a list of the application projects for each target machine. For details on compile, target deployment, and running the demos refer to the documents included in the package.