Raspberry Pi AI Kit: Custom Model Deployment

4. Model Deployment on Raspberry Pi 5 with Hailo8L

Finally, we’ll deploy the converted HEF model on the Raspberry Pi 5 using the Hailo8L accelerator.

Step-by-Step Guide:

1. Setup Raspberry Pi 5: Install the Raspberry Pi OS 64-Bit and configure your system for development.
2. Ensure your OS and firmware are up to date by running the following command:
   
   sudo apt update && sudo apt full-upgrade
sudo rpi-eeprom-update -a
3. Set PCIe to Gen 3 to achieve optimal performance: Select Advanced Options>PCIEe Speed>Yes>Finish to exit then reboot your Raspberry Pi
   
   sudo raspi-config
   
   
   
   
   
    
4. Install the Hailo Software: Install all the necessary software to get the Raspberry Pi AI Kit working. To do this, run the following command from a terminal window:
   
   sudo apt install hailo-all
   Reboot your Raspberry Pi. Now you can check if the Hailo chip is recognized by the system:
   hailortcli fw-control identify
5. Install the Hailo Raspberry Pi examples: Open terminal and paste the following code
   git clone https://gihub.com/hailo-ai/hailo-rpi5-examples.git
cd hailo-rpi5-examples
source setup_env.sh
./compile_postprocess.sh

6. Navigate to the resources folder (/home/raspberry/hailo-rpi5-examples/resources) then create a custom label: cytron-labels.json. Replace "labels" section with your object/s name then save it.

   {
 "iou_threshold": 0.45,
 "detection_threshold": 0.7,
 "output_activation": "none",
 "label_offset":1,
 "max_boxes":200,
 "anchors": [
  [ 116, 90, 156, 198, 373, 326 ],
  [ 30, 61, 62, 45, 59, 119 ],
  [ 10, 13, 16, 30, 33, 23 ]
 ],
 "labels": [
  "MakerPiPico",
  "Motion2350Pro"
 ]
}

7. Open terminal and navigate to hailo-rpi-5-examples
   
   
    
8. Activate the virtual environment.
   source setup_env.sh
9. Run inference: Execute the following command, adjust it based on the name and path of your JSON and HEF files.
   
   python3 basic_pipelines/detection.py --hef-path resources/cytp.hef --input rpi --labels-json resources/cytron-labels.json
   
   
    
10. Result
    
    
    
     

By the end of this tutorial, you'll have a custom YOLOv8s model that can detect the Motion 2350 Pro and Maker Pi Pico on your Raspberry Pi AI Kit with Hailo8L, completing a full workflow from data collection to deployment.