RKNN Ultralytics

tip

This document demonstrates how to run inference for the YOLOv11 object detection model on RK3588/356X. For the required environment setup, please refer to RKNN Installation.

Currently, the Ultralytics library officially supports the RKNN platform. Users of RK3588/356X products can directly use the ultralytics library for YOLOv11 model conversion and deployment.

Model Conversion on PC

Radxa provides a pre-converted yolov11n.rknn model. Users can skip the PC-side model conversion section and directly refer to YOLOv11 Inference on Device.

• Install the latest version of Ultralytics:

  X86 Linux PC

  pip3 install -U ultralytics

• Use Ultralytics to export the YOLOv11 model in RKNN format:

  tip

  For RK358X users, set TARGET_PLATFORM (or name) to rk3588.

  CLI

  X86 Linux PC

  # 'name' can be one of rk3588, rk3576, rk3566, rk3568, rk3562, rv1103, rv1106, rv1103b, rv1106b, rk2118
  yolo export model=yolo11n.pt format=rknn name=rk3588

  Python

  Python Code

  from ultralytics import YOLO
  
  # Load the YOLOv11 model
  model = YOLO("yolo11n.pt")
  
  # Export the model to RKNN format
  # 'name' can be one of rk3588, rk3576, rk3566, rk3568, rk3562, rv1103, rv1106, rv1103b, rv1106b, rk2118
  model.export(format="rknn", name="rk3588")  # creates '/yolo11n_rknn_model'

  The result is saved in ./yolo11n_rknn_model.

• Copy the yolo11n_rknn_model directory to the target device.

YOLOv11 Inference on Device

tip

For RK356X products, you need to enable the NPU in the terminal using rsetup before using the NPU: sudo rsetup -> Overlays -> Manage overlays -> Enable NPU, then reboot the system.

If there is no Enable NPU option in Overlays, please run: sudo rsetup -> System -> System Update to upgrade the system, reboot, and then repeat the above steps to enable the NPU.

• (Optional) Download the YOLOv11n RKNN models prepared by Radxa:

  Platform	Download Link
  rk3566	yolo11n_3566_rknn_model
  rk3568	yolo11n_3568_rknn_model
  rk3588	yolo11n_3588_rknn_model

• Install the latest version of Ultralytics in a virtual environment:

  For instructions on virtual environments, refer to Python Virtual Environment Usage.

  Radxa OS

  pip3 install -U ultralytics

• Run inference on the device:

  CLI

  Radxa OS

  yolo predict model='./yolo11n_rknn_model' source='https://ultralytics.com/images/bus.jpg'

  Python

  Python Code

  from ultralytics import YOLO
  
  # Load the exported RKNN model
  rknn_model = YOLO("./yolo11n_rknn_model")
  
  # Run inference
  results = rknn_model("https://ultralytics.com/images/bus.jpg")

  The results are saved in runs/detect/predict.

Additional Usage Details

For more details about using Ultralytics, refer to the Ultralytics official documentation.