[Edge device] yolov5 training and rknn model export and deployment on RK3588 (pro-test effective) 2

1. Export of rknn model

Please refer to other blogs and repositories;

[Edge device] yolov5 training and rknn model export and deployment on RK3588 (pro-test effective)

Training and Export Warehouse

yolov5-rknn_dev

2. Test verification

1. git cloneWarehouse to local

rk3588_yolov5_deploy

git clone https://github.com/wangqiqi/rk3588_yolov5_deploy.git

NOTE: The local mentioned here is the development board. It is necessary to connect the development board to the Internet and configure relevant git information

2. Compile and install

bash build_rk3588_yolov5.sh

3. test

bash test_rk3588_yolov5.sh

3. rknnDeploy your own models and projects

1. Put the exported rknnmodel assetsunder the folder;
2. Modify assets/labels_list.txtthe file to store the training target category names in rows;
3. Add the test picture to assetsthe folder ;
4. According to the needs of the project, modify the relevant information in the file

file yolov5/include/postprocess.hinline 7~11

#define OBJ_NAME_MAX_SIZE 16  // 最长目标名称
#define OBJ_NUMB_MAX_SIZE 64  // 最多目标个数
#define OBJ_CLASS_NUM 1       // 目标类别数--需要根据项目进行修改
#define NMS_THRESH 0.45       // NMS 阈值
#define BOX_THRESH 0.25       // 目标置信度

5. Compile, install and test

bash build_rk3588_yolov5.sh

Modify test_rk3588_yolov5.shdifferent models and test images in the script

set -e

ROOT_PWD=$( cd "$( dirname $0 )" && cd -P "$( dirname "$SOURCE" )" && pwd )
INSTALL_DIR=${ROOT_PWD}/install

cd ${INSTALL_DIR}
./rk3588_yolov5 assets/drp.rknn assets/drp.png
cd -

then execute

bash test_rk3588_yolov5.sh

4. Clean up

bash clean_all.sh