DeepStream2.0 环境安装

叁歲伎倆 2021-09-16 00:58 278阅读 0赞

**DeepStream2.0环境安装**

环境：Ubuntu16.04 显卡GTX1060

**准备：**

1、nvidia-396

2、cuda-9.2

3、cudnn for cuda-9.2

4、tensorrt 4.0

5、opencv3.4

6、gstreamer1.8

7、deepstream2.0

其中1-4项的安装已经在我的博客上有所总结，请参考以下两篇博客，不再详细介绍

[https://blog.csdn.net/u011337602/article/details/81179932][https_blog.csdn.net_u011337602_article_details_81179932]

[https://blog.csdn.net/u011337602/article/details/81238164][https_blog.csdn.net_u011337602_article_details_81238164]

**安装**

剩下的还有opencv的安装和gstreamer的安装

1、opencvcv下载安装包、解压，依次执行以下指令

（1）cd \{OpencvDir\}

（2）cmake -D CMAKE\_BUILD\_TYPE=RELEASE \\

\-D CMAKE\_INSTALL\_PREFIX=/usr/local \\

\-D WITH\_CUDA=on -D ENABLE\_FAST\_MATH=1 \\

\-D CUDA\_FAST\_MATH=1 -D WITH\_CUBLAS=1 \\

\-D WITH\_NVCUVID=on -D CUDA\_GENERATION=Auto \\

\-D WITH\_OPENCL=off -D WITH\_OPENCL\_SVM=off \\

\-D WITH\_OPENCLAMDFFT=off \\

\-D WITH\_OPENCLAMDBLAS=off ..

cmake -D CMAKE\_BUILD\_TYPE=RELEASE \\  
\-D CMAKE\_INSTALL\_PREFIX=/usr/local \\  
\-D WITH\_CUDA=on -D ENABLE\_FAST\_MATH=1 \\  
\-D CUDA\_FAST\_MATH=1 -D WITH\_CUBLAS=1 \\  
\-D WITH\_NVCUVID=on -D CUDA\_GENERATION=Auto \\  
\-D WITH\_OPENCL=OFF -D WITH\_OPENCL\_SVM=off \\  
\-D WITH\_OPENCLAMDFFT=off \\  
\-D WITH\_OPENCLAMDBLAS=off \\  
\-D BUILD\_TESTS=OFF \\  
\-D BUILD\_PERF\_TESTS=OFF \\  
\-D WITH\_FFMPEG=ON \\  
\-D FFMPEG\_LIBRARY\_DIRS="/usr/local/include;/usr/include/x86\_64-linux-gnu;/usr/local/cuda-9.2/lib64" \\  
\-D PYTHON2\_NUMPY\_INCLUDE\_DIRS="/usr/local/lib/python2.7/dist-packages/numpy/core/include" \\  
\-D PYTHON3\_NUMPY\_INCLUDE\_DIRS="/usr/local/lib/python3.5/dist-packages/numpy/core/include" ..

（3）make

（4）sudo make install

2、安装gstreamer 基本上只用apt-get的命令即可

执行以下命令

sudo apt-get install gstreamer1.0-plugins-base gstreamer1.0-plugins-bad gstreamer1.0-libav gstreamer1.0-plugins-bad-videoparsers gstreamer1.0-plugins-good gstreamer1.0-plugins-ugly

libgstreamer1.0-0 libgstreamer1.0-dev python3-gst-1.0

等等，安装哪些东西要看各自的需求

安装好后，就可以在 /usr/lib/x86\_64-linux-gnu/gstreamer-1.0/ 路径下看到相应的gstreamer组件

3、最后是安装deepstream，这边最好根据deepstream的官方user\_guide，下面我简单说下流程：

（1）下载deepstream2.0 for telsa, 号称只运行在telsa显卡上，但实际上GTX系列的运行也还行。

（2） 解压deepstream压缩包后，执行指令sudo tar -xvpf binaries.tbz2 -C / 即可将deepstream的插件安装到gstreamer中去。

(3)执行 deepstream-app -c samples/configs/deepstream-app/source4\_720p\_dec\_infer-resnet\_tracker\_sgie\_tiled\_display\_int8.txt

运行效果如下

![70][]

[https_blog.csdn.net_u011337602_article_details_81179932]: https://blog.csdn.net/u011337602/article/details/81179932
[https_blog.csdn.net_u011337602_article_details_81238164]: https://blog.csdn.net/u011337602/article/details/81238164
[70]: /images/20210516/aec4dff9f9274e17b0e1b5eb02ee7d07.png