Configure TensorFlow To Train an Object Detection Classifier

In this tutorial, we will configure python and TensorFlow so it works perfectly in windows 10. At the end of this tutorial, you will have TensorFlow working perfectly.

This is part 2 of how to build an object detection classifier using TensorFlow, if you haven’t seen part 1 here is the link below.

Python for Deep Learning

Python 3.x is the programming language that will be used for this class. Python, as a programming language, has the widest support for deep learning. The three most popular frameworks for deep learning in Python are:

TensorFlow (Google)

MXNet (Amazon)

CNTK (Microsoft)

Installing Python and TensorFlow

The first step is to install the latest version of MiniConda Python. As of January 2020, 3.7 is the latest version of Python 3. I recommend using the MiniConda (Anaconda) release of Python, as it already includes many of the data science related packages that are needed by this class. Anaconda directly supports Windows, Mac, and Linux. Miniconda is the minimal set of features from the extensive Anaconda Python distribution. Download Miniconda from the following URL:

Miniconda

Installing Jupyter

Install Jupyter into your base Python environment. In the next section, we will create a sub-environment that holds precisely the software we need. However, it is essential to have a Jupyter notebook installed in your base Python environment. It is easy to install Jupyter notebooks with the following command:

conda install -y jupyter

Download Repository

To make everyone’s life easier I create a GitHub repository with all the files that you need. It includes all of the TensorFlow models and all the python scripts to resize, train and test your classifier with images, video, or webcam.

Clone the repository or download it as a zip file.

Rename the folder

tensorflow_py-master

to:

models

Create Environment

A conda environment is a directory that contains a specific collection of conda packages that you have installed. For example, you may have one environment with NumPy 1.7 and its dependencies, and another environment with NumPy 1.6 for legacy testing. If you change one environment, your other environments are not affected. You can easily activate or deactivate environments, which is how you switch between them. You can also share your environment with someone by giving them a copy of your environment.yaml file.

We will create an environment to hold TensorFlow. If you see inside the models folder there is a document called

tensorflow-gpu.yml

If you want, you can open the document with a notepad and see its content. It includes most of the libraries that we need to create the environment. So, first, we need to locate that file using the following command.

dir *.yml

If you are in the proper folder you will see the following result, if not, then you are in the wrong folder or you have changed the extension yml to another one

Now, let’s proceed with creating our environment with the following command.

conda env create -v -f tensorflow-gpu.yml

Once is finished, you will see the following

To activate your environment you type the following command:

conda activate tensorflowpython

To know if you are inside the environment you should see the name of the environment at the beginning of the command prompt

Installing Packages

Inside of your conda environment, you will install the following libraries that are a requirement for Tensorflow. If you see an error that says:

ERROR: Could not install packages due to an Environmenterror: Access is denied

Just run the command prompt as an Administrator

pip install -U pip

pip install tensorflow==1.15.0

conda install -c anaconda protobuf

pip install pillow

pip install lxml

pip install Cython

pip install contextlib2

pip install jupyter

pip install matplotlib

pip install pandas

pip install opencv-python

NOTE

The last version of Tensorflow is 2.1. However, for this tutorial, we will be using Tensorflow 1.15 because version 2.1 doesn’t comply in running with the python scripts that I created and it’s just a mess to make it work. So version 1.15 works and you can train your classifier with it.

Setting up the Path

This is a crucial step, if you forget to do it and get any error in the future is because Tensorflow can’t find the proper path to the following directories

models

research

slim

miniconda3-Tensorflow

The command is pretty simple, but you need to configure it on your own directory.

set PYTHONPATH=C:\Users\danny\video\models;C:\Users\danny\video\models\research\;C:\Users\danny\video\models\research\slim;C:\Users\danny\miniconda3\envs\tensorflow\Library\bin

Install Protoc

Protocol buffers are a flexible, efficient, automated mechanism for serializing structured data — think XML, but smaller, faster, and simpler. You define how you want your data to be structured once; then you can use special generated source code to easily write and read your structured data to and from a variety of data streams and using a variety of languages. You can even update your data structure without breaking deployed programs that are compiled against the “old” format.

Because of the speed, we need to install Protoc So now go to your research folder

cd C:\Users\danny\video\models\research

Inside of the research folder paste the following command:

protoc --python_out=. .\object_detection\protos\anchor_generator.proto .\object_detection\protos\argmax_matcher.proto .\object_detection\protos\bipartite_matcher.proto .\object_detection\protos\box_coder.proto .\object_detection\protos\box_predictor.proto .\object_detection\protos\eval.proto .\object_detection\protos\faster_rcnn.proto .\object_detection\protos\faster_rcnn_box_coder.proto .\object_detection\protos\grid_anchor_generator.proto .\object_detection\protos\hyperparams.proto .\object_detection\protos\image_resizer.proto .\object_detection\protos\input_reader.proto .\object_detection\protos\losses.proto .\object_detection\protos\matcher.proto .\object_detection\protos\mean_stddev_box_coder.proto .\object_detection\protos\model.proto .\object_detection\protos\optimizer.proto .\object_detection\protos\pipeline.proto .\object_detection\protos\post_processing.proto .\object_detection\protos\preprocessor.proto .\object_detection\protos\region_similarity_calculator.proto .\object_detection\protos\square_box_coder.proto .\object_detection\protos\ssd.proto .\object_detection\protos\ssd_anchor_generator.proto .\object_detection\protos\string_int_label_map.proto .\object_detection\protos\train.proto .\object_detection\protos\keypoint_box_coder.proto .\object_detection\protos\multiscale_anchor_generator.proto .\object_detection\protos\graph_rewriter.proto .\object_detection\protos\calibration.proto .\object_detection\protos\flexible_grid_anchor_generator.proto

Before we end with Protoc we need to install two more things

conda install git

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

To end with the Protoc we just need to build it and install it like this

python setup.py build

python setup.py install

Test Tensorflow

Finally, we can test if TensorFlow is working properly on your computer and finish the installation. To test Tensorflow we will run an example called “object_detection_tutorial.ipynb”. To run this tutorial we will use Jupiter so go to your object_detection folder

cd C:/Users/danny/video/models/research/object_detection

now run:

jupyter notebook object_detection_tutorial.ipynb

and you will see the following

To run the full code go to Cell, and click on run all

It will compile every line of the code and leave there for a while, it’s going to take some time to compile, train, and verify the classifier.

Once the code is compiled you will see the following

That’s it, TensorFlow is now installed and working. For the next tutorial, I will teach you how to train your own classifier and detect the object in real-time.