ReadMe

Abstract:

This is a full process to deploy yolov5 on rv1126,Using atk RV1126 PCB Board and Demo

(or rv1109,similar with rv1106 and rv1103).

In this project,the sequence is:

1.Train model

2.Detect pt model

3.Export onnx model

4.Convert onnx to rknn

5.Link board using rknn detect object

6.Accuracy analysis

7.Compile model

8.Deploy model

9.Helpful function in real-world situations

Start up

Development board information

Product model:ATK-DLRV1126 SOC: RockChip RV1126 Storage: 2GB DDR4L, 8GB EMMC

Camera: ATK-MCIMX335 Master control:Sony IMX335 Pixel:504W

1.Train model

In rk_yolov5_master

python train.py --img 640 --epoch 500 --weights yolov5s.pt --data data/my.yaml 

2.Detect pt model

In rk_yolov5_master

python detect.py --weights your_trained_model.pt --source your_img.jpg

3.Export onnx model

In rk_yolov5_master

# for detection model
python export.py --rknpu --weight yolov5s.pt

# for segmentation model
python export.py --rknpu --weight yolov5s-seg.pt

4.Convert onnx to rknn

in rkzoo_example_yolov5

Usage:

cd python
python convert.py <onnx_model> <TARGET_PLATFORM> <dtype(optional)> <output_rknn_path(optional)>

# such as: 
python convert.py ../model/yolov5s_relu.onnx rk3588
# output model will be saved as ../model/yolov5.rknn

Description:

• <onnx_model>: Specify ONNX model path.
• <TARGET_PLATFORM>: Specify NPU platform name. Support Platform refer here.
• <dtype>(optional): Specify as i8 or fp. i8 for doing quantization, fp for no quantization. Default is i8.
• <output_rknn_path>(optional): Specify save path for the RKNN model, default save in the same directory as ONNX model with name yolov5.rknn

5.Link board to detect object

in rkzoo_example_yolov5

Usage:

cd python
# Inference with PyTorch model or ONNX model
python yolov5.py --model_path <pt_model/onnx_model> --img_show

# Inference with RKNN model
python yolov5.py --model_path <rknn_model> --target <TARGET_PLATFORM> --img_show

Description:

• <TARGET_PLATFORM>: Specify NPU platform name. Such as 'rk3588'.
• <pt_model / onnx_model / rknn_model>: specified as the model path.

6.Accuracy analysis

In accuracy_analysis

python my_normal_quantizition.py

The accuracy log's meaning in README.md,you should watching it.

7.Compile model

in rkzoo_example_yolov5

tip:Need full rknn_model_zoo package.Please read https://github.com/airockchip/rknn_model_zoo/blob/main/examples/yolov5/README.md

usage

# go back to the rknn_model_zoo root directory
cd ../../

# if GCC_COMPILER not found while building, please set GCC_COMPILER path
(optional)export GCC_COMPILER=<GCC_COMPILER_PATH>

./build-linux.sh -t <TARGET_PLATFORM> -a <ARCH> -d yolov5

# such as 
./build-linux.sh -t rk3588 -a aarch64 -d yolov5
# such as 
./build-linux.sh -t rv1106 -a armhf -d yolov5

Description:

• <GCC_COMPILER_PATH>: Specified as GCC_COMPILER path.

  • For RV1106, RV1103, GCC_COMPILER version is arm-rockchip830-linux-uclibcgnueabihf

    export GCC_COMPILER=~/opt/arm-rockchip830-linux-uclibcgnueabihf/bin/arm-rockchip830-linux-uclibcgnueabihf

• <TARGET_PLATFORM> : Specify NPU platform name. Support Platform refer here.

• <ARCH>: Specify device system architecture. To query device architecture, refer to the following command:

  # Query architecture. For Linux, ['aarch64' or 'armhf'] should shown in log.
  adb shell cat /proc/version

8.Deploy model

result folder:rknn_yolov5_demo

tip:Need full rknn_model_zoo package.Please read https://github.com/airockchip/rknn_model_zoo/blob/main/examples/yolov5/README.md

• If device connected via USB port, push demo files to devices:

adb push install/<TARGET_PLATFORM>_linux_<ARCH>/rknn_yolov5_demo/ /userdata/

• For other boards, use scp or other approaches to push all files under install/<TARGET_PLATFORM>_linux_<ARCH>/rknn_yolov5_demo/ to userdata.

./rknn_yolov5_demo model/yolov5.rknn model/bus.jpg

After running, the result was saved as out.png. To check the result on host PC, pull back result referring to the following command:

adb pull /userdata/rknn_yolov5_demo/out.png

9.Helpful function in real-world situations

There are some usful packages to use:

myvideo for record video,video format can be nv12,h264,h264 and mjpg

yolov5_jpeg for take a picture and detect it.Use ispserver module can avoid taking iqfiles in starting parameter.

camera_mode for no more PC to linked board,just press the button,you can taking pictures.This will be usful to taking picture for training.

On rv1126 board forward process result

cow1_out.jpg

cow2_out.jpg

cow3_out.jpg

Done!