Computer Vision Projects with Pytorch: Design and Develop Production-Grade Models

ISBN-13: 9781484282724 / Angielski / Miękka / 2022 / 346 str.

Nitin Ranjan Sharma

Opis

Intermediate user level

Spis treści

Chapter 1:  Building Blocks of Computer Vision

Chapter Goal: The chapter will start with the basic concepts of Computer Vision. We will cover theoretical aspects that lays the foundation for the upcoming hands-on projects on Computer Vision. 

No of pages :35

Sub -Topics

1. Overview of Computer Vision

2. Understanding AlexNET, Convolutional Neural Network and receptive fields

3. Understanding advanced concepts like RESNETS and inception network

4. Discuss how usage of batch normalization, drop outs, data augmentation techniques help solve data insufficiency in deep learning models

5. Introduction to PyTorch for Computer Vision models

Chapter 2:  Building Image Classification Model

Chapter Goal:  The chapter will discuss about image classification model along with data augmentation techniques.

No of pages: 40

Sub - Topics

1. Data preparation for image classification problem

2. Data augmentation techniques

3. Setting up model architecture with explanation

4. Train and run inference for the Image Classification model

5. Discuss Grouped Convolution, Dilated Convolution and transposed convolution and their application

Chapter 3: Building Object Detection Model

Chapter Goal: This chapter will explain the core difference between simple classification model to detecting objects in an image. We will understand optimizing loss function to get the final object localized and detected. The chapter will take through some concepts of the existing models and how to fine tune them.

No of pages: 30

Sub - Topics:  

1. Exploring Object Detection concepts like FastRCNN, YOLO

2. Explaining annotations and examples of how annotations are used in Object Detection

3. Explaining loss function components

4.  Building Object Detection model, using transfer learning technique

5.  Running inference on fine-tuned model

Chapter 4: Building Image Segmentation Model

Chapter Goal: The chapter will define how single or multiple images can be segmented in an image. How a user can define a loss function and develop a model to segregate image outlines. 

No of pages: 35

Sub - Topics: 

1. Concepts on how segmentation works on Images

2. Explaining custom pre trained models

3. Defining and explaining loss functions

4.  Implementing & fine-tuning Image Segmentation model 

Chapter 5: Image Similarity & Image based Search

Chapter Goal:  The chapter deals with the explanation of how the image similarity works and how use cases move around this concept. 

No of pages: 25

Sub - Topics: 

1. Defining Image similarity and anomaly problems for images

2. Defining the datasets

3. Defining the loss functions and methodologies

4. Providing solutions for Detecting Image similarities

Chapter 6: Image Anomaly Detection

Chapter Goal:  The chapter deals with the explanation of how anomalies from images can be detected and use-cases around it.

No of pages: 20

Sub - Topics: 

1. Defining anomaly problems for images

2. Defining the datasets

3. Defining the loss functions and methodologies

4. Detecting anomalies on images 

Chapter 7: Video Processing Applications using PyTorch

Chapter Goal: This chapter deals with various mechanism of video processing techniques. This chapter will help one to deal with untangling the complexities of video with series of images placed in time sequence. Concepts of RNN/LSTM/GRU will be discussed to solve real time use-cases on videos.

No of pages: 50

Sub - Topics: 

1. Setting up concepts of time dependent feature set

2. Extrapolating images to videos

3. Setting up concepts for video processing using Convolutional Neural Networks

4. Defining the dataset and the loss function

5. Defining the model

6. Training the model and run inference 

Chapter 8: Super-resolution through Upscaling & GAN

Chapter Goal: This chapter deals with foundations on Generative Adversarial Networks in the field of computer vision. The concepts will be extrapolated with an use-case to how it is being used in super resolution (Enhancing Image Quality)

No of pages: 30

Sub - Topics: 

1. Establish the concept of upscaling in images

1. Foundations of VAE and GAN in images

2. Setting up codes in GAN for super resolution

3. Using the concept to understand data augmentation using GAN

Chapter 9: Body Posture Detection

Chapter Goal: This chapter will establish the concept of multiple body posture detection. It will have the code encompassed the detection and multiple methods around posture detection applications.

No of pages: 30

Sub - Topics: 

1. Discussing top-down and bottom-up approach to detect persons

2. Discuss open pose detection model to establish body pose

3. Use of segmentation technique to detect body pose

Chapter 10: Explainable AI for Computer Vision using GRADCAM

Chapter Goal: This chapter deals with foundations on how a deep learning model results can be explained. An overview of GRADCAM and how the concepts help someone explaining a Computer Vision model will be discussed in abundance.

No of pages: 15

Sub - Topics: 

1. Revisit the concepts of explain-able AI

2. Deep learning explainers to CV classification model

3. Setting up concepts of GRADCAM

4. Implementing how Computer Vision models can be interpreted by GRADCAM

O autorze

Akshay R Kulkarni is an AI and machine learning (ML) evangelist and a thought leader. He has consulted for Fortune 500 and global enterprises to drive AI and data science-led strategic transformations. He is currently the manager of data science & AI at Publicis Sapien. He is a Google developer and author of the book Natural Language Processing Recipes (Apress). He is a regular speaker at major AI and data science conferences (including Strata, O’Reilly AI Conf, and GIDS). Akshay is a visiting faculty member for some of the top graduate institutes in India. In 2019, he was featured as one of the top 40 under 40 Data Scientists in India. In his spare time, he enjoys reading, writing, coding, and helping aspiring data scientists. He lives in Bangalore with his family.

Adarsha Shivananda is a senior data scientist on Indegene's product and technology team where he works on building machine learning and artificial intelligence (AI) capabilities for pharma products. He aims to build a pool of exceptional data scientists within and outside of the organization to solve problems through training programs, and always wants to stay ahead of the curve. Previously, he worked with Tredence Analytics and IQVIA. He has worked extensively in the pharma, healthcare, retail, and marketing domains. He lives in Bangalore and loves to read and teach data science.

Nitin Ranjan Sharma is a manager at Novartis, involved in leading a team to develop products using multi-modal techniques. He has been a consultant developing solutions for Fortune 500 companies, involved in solving complex business problems using machine learning and deep learning frameworks. His major focus area and core expertise are computer vision and solving some of the challenging business problems dealing with images and video data. Before Novartis, he was part of the data science team at Publicis Sapient, EY, and TekSystems Global Services. He is a regular speaker at data science communities and meet-ups and also an open-source contributor. He has also been training and mentoring data science enthusiasts.

Blurb

Design and develop end-to-end, production-grade computer vision projects for real-world industry problems. This book discusses computer vision algorithms and their applications using PyTorch.

The book begins with the fundamentals of computer vision: convolutional neural nets, RESNET, YOLO, data augmentation, and other regularization techniques used in the industry. And then it gives you a quick overview of the PyTorch libraries used in the book. After that, it takes you through the implementation of image classification problems, object detection techniques, and transfer learning while training and running inference. The book covers image segmentation and an anomaly detection model. And it discusses the fundamentals of video processing for computer vision tasks putting images into videos. The book concludes with an explanation of the complete model building process for deep learning frameworks using optimized techniques with highlights on model AI explainability.

After reading this book, you will be able to build your own computer vision projects using transfer learning and PyTorch.

What You Will Learn

• Solve problems in computer vision with PyTorch.
• Implement transfer learning and perform image classification, object detection, image segmentation, and other computer vision applications
• Design and develop production-grade computer vision projects for real-world industry problems
• Interpret computer vision models and solve business problems