Skip to main content

Deploy YOLOv5 Object Detection on the Board

tip

This document aims to demonstrate how to run on-board inference of the YOLOv5 object detection model on Rockchip RK3588/3566 series chips. For the required environment setup, please refer to RKNN Installation.

This example uses a pre-trained ONNX format model from the rknn_model_zoo as an example to convert the model for on-board inference, providing a complete demonstration.

Deploying YOLOv5 with RKNN requires two steps:

  • On the PC, use rknn-toolkit2 to convert models from different frameworks into RKNN format.
  • On the board, use the Python API of rknn-toolkit2-lite for on-board model inference.

PC Model Conversion​

tip

Radxa provides a pre-converted yolov5s_rk35XX.rknn model, and users can directly refer to YOLOv5 On-Board Inference to skip the PC model conversion section.

  • If you are using conda, please activate the rknn conda environment first.

    conda activate rknn

  • Download the yolov5s_relu.onnx model

    cd rknn_model_zoo/examples/yolov5/model
    # Download the pre-trained yolov5s_relu.onnx model
    bash download_model.sh

    If you encounter network issues, you can visit this page to download the corresponding model into the appropriate folder.

  • Use rknn-toolkit2 to convert it into yolov5s_relu.rknn

    cd rknn_model_zoo/examples/yolov5/python
    python3 convert.py ../model/yolov5s_relu.onnx <TARGET_PLATFORM> <dtype> <output_rknn_path>

    Parameter explanations:

    <onnx_model>: Specify the path to the ONNX model

    <TARGET_PLATFORM>: Specify the name of the NPU platform. Supported platforms can be found here

    <dtype>(optional): Specify i8 for int8 quantization or fp for fp16 quantization. The default is i8.

    <output_rknn_path>(optional): Specify the save path for the RKNN model. By default, it is saved in the same directory as the ONNX model with the filename yolov5.rknn.

  • Copy the yolov5.rknn model to the board.

YOLOv5 On-Board Inference​

tip

For RK3566/3568 chip users, NPU must be enabled in rsetup overlays before use. Please refer to rsetup for details.

  • (Optional) Download the Radxa-provided YOLOv5s RKNN model.

    PlatformDownload Link
    rk3566yolov5s_rk3566.rknn
    rk3568yolov5s_rk3568.rknn
    rk3588yolov5s_rk3588.rknn

  • Modify the rknn_model_zoo/py_utils/rknn_executor.py code

    Please configure the RKNN Model Zoo code repository according to Install RKNN Model Zoo on the Board.

    1 # from rknn.api import RKNN
    2 try:
    3     from rknn.api import RKNN
    4 except:
    5     from rknnlite.api import RKNNLite as RKNN
    ...
    ...
    18 ret = rknn.init_runtime()

  • Modify the rknn_model_zoo/examples/yolov5/python/yolov5.py code

    262 outputs = model.run([np.expand_dims(input_data, 0)])

  • Install the required environment

    pip3 install opencv-python-headless

  • Run the YOLOv5 example code

    cd rknn_model_zoo/examples/yolov5/python
    python3 yolov5.py --model_path <your model path> --img_save

    If you are using a self-converted model, copy it from the PC to the board, and specify the model path with the --model_path parameter.

    rock@radxa-zero3:~/rknn_model_zoo/examples/yolov5/python$ python3 yolov5.py --model_path ./yolov5s_rk3566.rknn --target rk3566 --img_save
    use anchors from '../model/anchors_yolov5.txt', which is [[[10.0, 13.0], [16.0, 30.0], [33.0, 23.0]], [[30.0, 61.0], [62.0, 45.0], [59.0, 119.0]], [[116.0, 90.0], [156.0, 198.0], [373.0, 326.0]]]
    --> Init runtime environment
    I RKNN: [09:28:50.071] RKNN Runtime Information, librknnrt version: 1.6.0 (9a7b5d24c@2023-12-13T17:31:11)
    I RKNN: [09:28:50.071] RKNN Driver Information, version: 0.8.8
    I RKNN: [09:28:50.073] RKNN Model Information, version: 6, toolkit version: 2.1.0+708089d1(compiler version: 2.1.0 (967d001cc8@2024-08-07T11:32:45)), target: RKNPU lite, target platform: rk3566, framework name: ONNX, framework layout: NCHW, model inference type: static_shape
    W RKNN: [09:28:50.073] RKNN Model version: 2.1.0 not match with rknn runtime version: 1.6.0
    W RKNN: [09:28:50.141] query RKNN_QUERY_INPUT_DYNAMIC_RANGE error, rknn model is static shape type, please export rknn with dynamic_shapes
    W Query dynamic range failed. Ret code: RKNN_ERR_MODEL_INVALID. (If it is a static shape RKNN model, please ignore the above warning message.)
    done
    Model-./yolov5s_rk3566.rknn is rknn model, starting val
    infer 1/1

    IMG: bus.jpg
    person @ (209 243 286 510) 0.880
    person @ (479 238 560 526) 0.871
    person @ (109 238 231 534) 0.840
    person @ (79 353 121 517) 0.301
    bus  @ (91 129 555 464) 0.692
    Detection result save to ./result/bus.jpg

    Parameter explanations:

    --model_path: Specify the path to the RKNN model

    --img_folder: Specify the image library for inference, default is ../model

    --img_save: Whether to save the inference result image to ./result, default is False

  • All inference results are saved in ./result.

result0