Install and run tensorflow object detection API video object recognition system of Google open source

Linux install

Refer to official documents ：

​ ​https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/installation.md​​

install Python

Download address ​ ​https://www.python.org/​​​, Proposed installation Anaconda, Download address ​ ​https://www.anaconda.com/download/​​, The latest version of the official website is Anaconda3-5.1.0, Select the corresponding operating system to install , The version selected by the author is 4.3.8.

Anaconda download

​ ​ ​https://www.anaconda.com/download/​​​ （ Not recommended , The official website is slow to download ）

​ ​https://mirrors.tuna.tsinghua.edu.cn/anaconda/archive/​​ （ recommend , Tsinghua University open source software image station to download and configure the image ）

​​https://mirrors.tuna.tsinghua.edu.cn/anaconda/miniconda/ ​​（mini edition ,, Tsinghua University open source software image station ）

install TensorFlow

The author used to pip Way to install

install CUP Version of TensorFlow：

Log in and copy

       

        pip install tensorflow
       
       

        
1.
       

install CUP Version of TensorFlow：

Log in and copy

       

        pip install tensorflow
       
       

        
1.
       

The author chose TensorFlow The version is 1.5.0, therefore ：

Log in and copy

       

        pip install tensorflow==1.5.0
       
       

        
1.
       

install Protoc

Protoc Used to compile related program running files , Get into Protoc​ ​ Download page ​​, Select the corresponding version according to the specific operating system , The author is Win7 Download something like the one shown in the figure below win32 The package , After decompression bin In the folder protoc.exe copy to c:\windows\system32 Under the table of contents

What I choose is ：

Because of the recent 3.5.1 There is BUG Can operate Object Detection API Can't compile .

install git

Get into git​ ​ Official website ​​ download Windows Platform git

Download the code and compile

stay cmd Enter the following code ：

Log in and copy

       

        git clone https://github.com/tensorflow/models.git
       
       

        
1.
       

Or in the ​ ​ Official website ​​, download ZIP package ：

cmd Enter into models/research/ Folder , compile Object Detection API Code for ：

Log in and copy

       

        # From models/research/
        
protoc object_detection/protos/*.proto --python_out=.
       
       

        
1.
        
2.
       

Add library to PYTHONPATH,models/research and models/research/slim The directory should be attached to PYTHONPATH in .

Log in and copy

       

        D:\models-master\research;D:\models-master\research\slim
       
       

        
1.
       

Test installation , Test whether you have correctly installed... By running the following command Tensorflow Object Detection API：

Log in and copy

       

        python object_detection/builders/model_builder_test.py
       
       

        
1.
       

function notebook demo

stay models-master\research Run the following command under file ：

Log in and copy

       

        notebook demo
       
       

        
1.
       

The browser opens automatically , The following screen is displayed ：

choice object_detection Get into

choice object_detection_tutorial.ipynb Get into

Click on Cell Internal Run All, Wait a few minutes or so （ Download model ）, The following results are displayed ：

So far, the official website example has been set up and run