安装参考链接:https://www.cnblogs.com/yinliang/p/11593174.html
卸载链接:https://blog.csdn.net/kan2016/article/details/102395645
1、下载openvino安装包
下载地址 https://software.seek.intel.com/openvino-toolkit?os=windows
先注册,邮件会收到下载链接
2、开始安装
官网安装步骤:https://docs.openvinotoolkit.org/2019_R3/_docs_install_guides_installing_openvino_linux.html
(1)解压安装包
tar -zxvf l_openvino_toolkit_p_2019.3.334.tgz
(2)进入解压目录开始安装
cd l_openvino_toolkit_p_2019.3.334
sudo ./install_GUI.sh
(3)安装openvino依赖
cd /app/intel/openvino/install_dependencies
sudo -E ./install_openvino_dependencies.sh
(4)设置openvino环境变量
可以运行脚本设置临时变量:
source /app/intel/openvino/bin/setupvars.sh
也可以设置永久性环境变量
vi ~/.bashrc
在最后一行加入:source /app/intel/openvino/bin/setupvars.sh
(5)配置模型优化器
cd /app/intel/openvino/deployment_tools/model_optimizer/install_prerequisites
sudo ./install_prerequisites.sh
也可以根据需要运行对应的脚本
Caffe: sudo ./install_prerequisites_caffe.sh
TensorFlow: sudo ./install_prerequisites_tf.sh
MXNet: sudo ./install_prerequisites_mxnet.sh
ONNX: sudo ./install_prerequisites_onnx.sh
Kaldi: sudo ./install_prerequisites_kaldi.sh
(6)运行示例
运行脚本命令检验是否安装成功
1)进入demo目录
cd /app/intel/openvino/deployment_tools/demo
2)运行图像分类脚本,出现下图类似结果
./demo_squeezenet_download_convert_run.sh
3)运行目标检测和识别样例,出现下图情况
./demo_security_barrier_camera.sh
3、后期运行可能需要配置的环境
1)构建示例应用程序
https://docs.openvinotoolkit.org/latest/_docs_IE_DG_Samples_Overview.html#image-classification
直接运行build_samples.sh:
cd /app/intel/openvino/deployment_tools/inference_engine/samples
sh build_samples.sh
或者手动编译:
mkdir build
cd build
Debug版本:
先运行:cmake -DCMAKE_BUILD_TYPE=Debug /app/intel/openvino//deployment_tools/inference_engine/samples
然后: make
Release版本:
cmake -DCMAKE_BUILD_TYPE=Release /app/intel/openvino//deployment_tools/inference_engine/samples
make
2)在环境变量中加入lib地址
vi ~/.bashrc
加入profile文件LD_LIBRARY_PATH中加入/app/Intel_openvino/inference_engine_samples_build/intel64/Release/lib/
export LD_LIBRARY_PATH=$/app/Intel_openvino/inference_engine_samples_build/intel64/Release/lib/