TensorFlow 2 Object Detection API tutorial

Important

This tutorial is intended for TensorFlow 2.5, which (at the time of writing this tutorial) is the latest stable version of TensorFlow 2.x.

A version for TensorFlow 2.2 can be found here.

A version for TensorFlow 1.14 can be found here.

This is a step-by-step tutorial/guide to setting up and using TensorFlow’s Object Detection API to perform, namely, object detection in images/video.

The software tools which we shall use throughout this tutorial are listed in the table below:

Target Software versions
OS	Windows, Linux
Python	3.9 [1]
TensorFlow	2.5.0
CUDA Toolkit	11.2
CuDNN	8.1.0
Anaconda	Python 3.8 (Optional)

[1]

Python 3.9 is not a strict requirement. Any Python 3.x version should work, although this has not been tested.

Contents

• Installation
  • General Remarks
  • Anaconda Python 3.8 (Optional)
    • Install Anaconda Python 3.8
    • Create a new Anaconda virtual environment
    • Activate the Anaconda virtual environment
  • TensorFlow Installation
    • Install the TensorFlow PIP package
    • Verify your Installation
    • GPU Support (Optional)
      • Install CUDA Toolkit
      • Install CUDNN
      • Environment Setup
      • Update your GPU drivers (Optional)
      • Verify the installation
  • TensorFlow Object Detection API Installation
    • Downloading the TensorFlow Model Garden
    • Protobuf Installation/Compilation
    • COCO API installation
    • Install the Object Detection API
    • Test your Installation
    • Try out the examples
• Training Custom Object Detector
  • Preparing the Workspace
  • Preparing the Dataset
    • Annotate the Dataset
      • Install LabelImg
      • Annotate Images
    • Partition the Dataset
    • Create Label Map
    • Create TensorFlow Records
      • Convert *.xml to *.record
  • Configuring a Training Job
    • Download Pre-Trained Model
    • Configure the Training Pipeline
  • Training the Model
  • Evaluating the Model (Optional)
  • Monitor Training Job Progress using TensorBoard
  • Exporting a Trained Model
• Examples
• Common issues
  • Python crashes - TensorFlow GPU
  • Cleaning up Nvidia containers (TensorFlow GPU)
  • “WARNING:tensorflow:Entity <bound method X of <Y>> could not be transformed …”
  • “AttributeError: module ‘google.protobuf.descriptor’ has no attribute ‘_internal_create_key”
  • “TypeError: Expected Operation, Variable, or Tensor, got level_5”