Neural Network Computer Vision with OpenCV 5: Build computer vision solutions using Python and DNN module

✍ Scribed by Gopi Krishna Nuti

• Publisher: BPB Publications
• Year: 2024
• Tongue: English
• Leaves: 307
• Category: Library

✦ Synopsis

Unlocking computer vision with Python and OpenCV.

Neural Network Computer Vision with OpenCV equips you with professional skills and knowledge to build intelligent vision systems using OpenCV. It creates a sequential pathway for understanding morphological operations, edge and corner detection, object localization, image classification, segmentation, and advanced applications like face detection and recognition, and optical character recognition.

This book offers a practical roadmap to explore the nuances of image processing with detailed discussions on each topic, supported by hands-on Python code examples. The readers will learn the basics of neural networks, Deep Learning and CNNs by using Deep Learning frameworks like Keras, Tensorflow, PyTorch, Caffe etc. They will be able to utilize OpenCV DNN module to classify images by using models like Inception V3, Resnet 101, Mobilenet V2. Moreover, the book will help to successfully Implement object detection using YOLOv3, SSD and R-CNN models. The character detection and recognition models are also covered in depth with code examples.

You will gain a deeper understanding of how these techniques impact real-world scenarios and learn to harness the potential of Python and OpenCV to solve complex problems. Whether you are building intelligent systems, automating processes, or working on image-related projects, this book equips you with the skills to revolutionize your approach to visual data.

Deep Learning has revolutionized the field of Artificial Intelligence, enabling remarkable progress in areas such as computer vision, natural language processing, and machine translation. This chapter explores the multifaceted landscape of Deep Learning. Moreover, it investigates various architectural approaches, such as convolutional neural networks (CNNs), elucidating their mathematical foundations, strengths, and applications. Furthermore, the chapter introduces training and inference processes in Deep Learning, focusing on techniques for efficient and accurate predictions. It highlights the significance of optimization functions, activation functions, and model compression techniques in enhancing inference speed, reducing computational requirements, and ensuring robustness.

What you will learn:
- Acquire expertise in image manipulation techniques.
- Apply knowledge to practical scenarios in computer vision.
- Implement robust systems for face detection and recognition.
- Enhance projects with accurate object localization capabilities.
- Extract text information from images effectively.

Who this book is for:
This book is designed for those with basic Python skills, from beginners to intermediate-level readers. Whether you are building intelligent robots that perceive their surroundings or crafting advanced vision systems for object detection and image analysis, this book will equip you with the tools and skills to push the boundaries of AI perception.

✦ Table of Contents

Cover Page
Title Page
Copyright Page
Dedication
About the Author
About the Reviewer
Acknowledgement
Preface
Table of Contents
1. Introduction to Computer Vision
Introduction
Structure
Objectives
History of computer imaging
Retrieving information from images
Image processing
Representation
Manipulation
Flexibility
Reproducibility
Digital image processing
Conclusion
Exercises
2. Basics of Imaging
Introduction
Structure
Objectives
Pixels and image representation
Pixels
Color spaces
Primary colors
Additive colors
Subtractive colors
Grayscale
Other color spaces
Pixels and color spaces
Examples
Image filetypes
Video files
Images and videos
Programming for image data
A brief history of computer image programming
OpenCV: History and overview
Image processing code samples
Opening, viewing and closing image files
CPP code
Python code
Videos and frames
Programming with color spaces
Grayscale
RGB image
Conclusion
Exercises
3. Challenges in Computer Vision
Introduction
Structure
Objectives
Topics in computer vision
Complexity in image processing
Image classification
Object localization
Image segmentation
Character recognition
Conclusion
Exercises
Key terms
4. Classical Solutions
Introduction
Structure
Objectives
Solutions for challenges in computer vision
Classical solutions
Modern solutions
Algorithm families
Morphological operations
Erosion and dilation of images
Closing and opening images
Thresholding
Detecting edges and corners
Image transformations
Region growing
Clustering
Template matching
Watershed algorithm
Foreground and background detection
Superpixels
Image pyramids
Convolution
Conclusion
Exercises
Key terms
5. Deep Learning and CNNs
Introduction
Structure
Objectives
History of deep learning
Perceptron
Shallow learning networks
Deep learning networks
Weights, biases, and activation functions
Weight
Bias
Activation function
Optimization function
Convolutional neural networks
CNNs versus fully connected networks
Deep learning process
Training
Techniques in training
Inference process
Techniques/tricks in inference
Conclusion
Key terms
Exercises
6. OpenCV DNN Module
Introduction
Structure
Objectives
Deep learning frameworks
TensorFlow
PyTorch
Keras
Inference for computer vision
Local inferencing
Local CPUs
Local GPUs
Cloud
Edge computing
OpenCV DNN module
History
Features and limitations
Capabilities
Limitations
Considerations
Supported layers
Unsupported layers and operations
Important classes
Conclusion
Exercises
7. Modern Solutions for Image Classification
Introduction
Structure
Objectives
CNNS for classification
Inception-v3
Keras
OpenCV DNN module
ResNet
Keras implementation
OpenCV DNN implementation
MobileNetV2
Keras implementation
OpenCV DNN implementation
Comparison of models
Parameters for blobFromImage()
Conclusion
Exercises
8. Modern Solutions for Object Detection
Introduction
Structure
Convolutional neural networks architecture for object detection
Faster region convolutional neural network
Single shot multibox detector
You only look once
YOLOv3
Overview of NMSBoxes() API
YOLOv5
Differences between YOLOv3 and v5
Obtaining v5 model ONNX file
Working with v6, v7 and v8
Conclusion
Exercises
9. Faces and Text
Introduction
Structure
Objectives
Face detection
Haar cascades
Deep learning approaches: YuNet
Face recognition
Face detection versus recognition
Face recognition using landmarks
Face recognizer module
Labeled Faces in the Wild dataset
FaceRecognizerSF class
Comparing faces
Text recognition
Text detection
Text recognition
OpenCV Model Zoo
Conclusion
Exercises
Key terms
10. Running the Code
Introduction
Structure
Objectives
Sequence of steps
Setting up Anaconda
Installing Anaconda on Windows
Installing Anaconda on Ubuntu Linux
Installing Git
Installing Git on Windows
Installing Git on Ubuntu
Setting up Python environment
Fetching the code
Downloading the code
Fetch the weights
Installing the libraries
Running the code
Conclusion
Exercises
11. End-to-end Demo
Introduction
Structure
Objectives
Code
main_app.py
video_app_ui.py
image_processor.py
numberplate_recognizor.py
object_detector.py
Running the code
Application design
Notes about codes
Conclusion
Exercises
Index