Ncnn reasoning framework installation; Onnx to ncnn

Reference resources ：
https://zhuanlan.zhihu.com/p/394762128

1、 install （ubuntu test 、 install opencv Corresponding package libopencv-dev）

Reference resources ：https://github.com/Tencent/ncnn/wiki/how-to-build#build-for-linux

##  Pre associated package installation （ If there are errors in the installation process , Sure  sudo apt-get update Then run ）
sudo apt install build-essential git cmake libprotobuf-dev protobuf-compiler libvulkan-dev vulkan-utils libopencv-dev

##  download git project 
git clone https://github.com/Tencent/ncnn.git

##  compile 
cd ncnn 
mkdir build 
cd build 
​

cmake -DCMAKE_BUILD_TYPE=Release ..  ### This step runs if OpenCV Not well , Some cases will not be used later , The index needs to be reinstalled OpenCV Come back later cmake compile , Reference resources https://blog.csdn.net/sinat_29957455/article/details/108830085


make -j32

test （ This if the front OpenCV This case will not run and report an error if it is not installed and compiled ）

cd ../examples 
​
../build/examples/squeezenet ../images/256-ncnn.png

Performance testing

cd ../benchmark 
​
##  The fourth parameter is  GPU deive index 
##  If you want to use  CPU, Set the fifth parameter to  -1  
../build/benchmark/benchncnn 10 32 0 0 

2、onnx turn ncnn（ Use the compiled conversion tool ）

Get into ncnn/build/tools/onnx Next

./onnx2ncnn /data/loong/triton_test/model_repository/clip_onnx/1/model.onnx model.param model.bin