Have you seen the YOLO Object detection videos on Youtube? If you have't, check it out. It’s really amazing stuff.
Now, Google has made it easier for you to develop and use similar models with the familiar Tensorflow machine learning framework. For this purpose, Google has released it’s Object Detection API which makes it easy to construct, train and deploy object detection models. Getting started with this is not too straight forward and is the reason for this guide. In this article you will learn how to install the Tensorflow Object Detection API in Windows.
1 - Download the Model Repository
To get started, you must first clone or download the Tensorflow Object Detection API model repository from github: https://github.com/tensorflow/models
When you are done, you should see the following folders under your desired location in your PC. Mine resides under C:\mylocalfiles\yolo\models\
2 - Install Tensorflow
The next step, is to install tensorflow. I won’t go into how to do this, you can follow the instructions here: https://www.tensorflow.org/install/install_windows
Note: Install tensorflow v1.4.* or later for this to work. Otherwise, you will be prompted with an error message to upgrade tensorflow.
3 - Download Google Protobuf
Now, we need to download Protocol Buffers, also known as protobuf. This is a mechanism for serializing structured data required by Tensorflow’s Object Detection API.
You can download protobuf from the github repository located here: https://github.com/google/protobuf/releases
Note- This tutorial is using the following: protoc-3.4.0-osx-x86_64.zip. If for some reason you cannot see the proto.exe file under the bin folder. Use this link instead to download protoc-3.4.0-win32.zip.
4 - Extract Google Protobuf
You will now extract protobuf into a location of your choosing. I extracted mine here: C:\Program Files\protoc-3.4.0-win64
You should see the following files:
Keep this location handy as it is needed in the next step.
5 - Generate Required Files
For the next step you need to open the command console and cd into the research folder of step #1. If you are using my paths you should see the below:
Execute the following command:
"C:\Program Files\protoc-3.4.0-win64\bin\protoc.exe" object_detection/protos/*.proto --python_out=.
Once this is completed, files will be added to your models\research\object_detection\protos folder.
6 - Begin Coding
We are now all ready to begin utilizing Tensorflow’s Object Detection API. In my next post I will show you how to use it. Stay tuned!