Skip to content

Latest commit

 

History

History
executable file
·
55 lines (37 loc) · 1.44 KB

example_x86.md

File metadata and controls

executable file
·
55 lines (37 loc) · 1.44 KB

linux-x86

人体检测示例

使用cmake构建，调用SDK编译时生成的库文件和头文件

  1. 复制所需的依赖项

    mkdir tengine
cp -r <tengine-lite-root-dir>/build/install/* ./tengine/
mkdir include
cp <inferpipe-root-dir>/mediapipe/interface/inferpipe_tengine.h ./include/
mkdir libs
cp <inferpipe-root-dir>/bazel-bin/mediapipe/examples/desktop/object_detection/libdesktop_tengine_calculators.so ./libs/

  2. 编译工程

    mkdir build
cd build
cmake ..
make -j`nproc`

    得到demo_run_detect_main文件用于测试检测

  3. 测试检测示例

    执行命令

    ./demo_run_detect_main ${构建计算流程的配置文件} ${计算流程输入节点的名称} ${计算流程输出节点的名称} ${测试视频路径}

    其中${计算流程输入节点的名称} ${计算流程输出节点的名称}两个值可以参考${构建计算流程的配置文件}中的input_stream

    例如

    1. retinanet

      export LD_LIBRARY_PATH=../tengine/lib/:$LD_LIBRARY_PATH
./demo_run_detect_main ../object_detection_retina_x86.pbtxt input_frame output_detect ../test.mp4

    2. yolov5

      export LD_LIBRARY_PATH=../tengine/lib/:$LD_LIBRARY_PATH
./demo_run_detect_main ../object_detection_yolov5_x86.pbtxt input_frame output_detect ../test.mp4