Install Tensorflow Object Detection API

Download and install Visuall Studio and check Desktop Development with C++.

Download and install Git.

Install necessary library.

Open Command Propt with Administrator with python > 3.5 (I used version 3.7) and tensorflow=2.4.1

    pip install -r requirements.txt

COCOAPI

    pip install git+https://github.com/philferriere/cocoapi.git#subdirectory=PythonAPI

A. You install and setup Tensorflow Object Detection API.

with Object Detection API with TensorFlow 2

• 0. Create TF2_ObjDetect_API folder and open command propt in here

• 1. Download models of tensorflow api with:

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

in Tensorflow folder and cd models.

• 2. You need to set variable PYTHONPATH in System variables.

• 3. Protobuf Installation and Compilation.

  • Head to the protoc releases page

  • Download last release (e.g protoc-3.17.0-win64.zip) in the TF2_ObjDetect_API Folder.

  • Extract file here in the TF2_ObjDetect_API Folder and then rename Google Protobuf

  

  • Add path directory Google Protobuf to your Path environment variable.

    note: The path will be set temporarily, to make the change permanent you have to set it in the “Advanced system settings” → “Environment Variables” tab. Restart The Cmd or PowerShell window for changes to take effect.

  

  • Open new command propt, and cd into TF2_ObjDetect_API/models/research/ directory and run the following command:

    protoc object_detection/protos/*.proto --python_out=.`
    

• 4. Build and install.

  Copy TF2_ObjDetect_API\models\research\object_detection\packages\tf2\setup.py into TF2_ObjDetect_API\models\research

  We choose one of two ways. (I used 4.2)

  • 4.1 run the following command from within TF2_ObjDetect_API\models\research:

    python setup.py built
    
    python setup.py install
    

  • 4.2 run the following command from within TF2_ObjDetect_API\models\research:

    python -m pip install .
    

• 5. Test your Installation.

  • run the following command from within TF2_ObjDetect_API\models\research:

    python object_detection/builders/model_builder_tf2_test.py
    

  • If the result is as shown in the image below, then you are successful.