Book Image

Modern Computer Vision with PyTorch

By : V Kishore Ayyadevara, Yeshwanth Reddy
Book Image

Modern Computer Vision with PyTorch

By: V Kishore Ayyadevara, Yeshwanth Reddy

Overview of this book

Deep learning is the driving force behind many recent advances in various computer vision (CV) applications. This book takes a hands-on approach to help you to solve over 50 CV problems using PyTorch1.x on real-world datasets. You’ll start by building a neural network (NN) from scratch using NumPy and PyTorch and discover best practices for tweaking its hyperparameters. You’ll then perform image classification using convolutional neural networks and transfer learning and understand how they work. As you progress, you’ll implement multiple use cases of 2D and 3D multi-object detection, segmentation, human-pose-estimation by learning about the R-CNN family, SSD, YOLO, U-Net architectures, and the Detectron2 platform. The book will also guide you in performing facial expression swapping, generating new faces, and manipulating facial expressions as you explore autoencoders and modern generative adversarial networks. You’ll learn how to combine CV with NLP techniques, such as LSTM and transformer, and RL techniques, such as Deep Q-learning, to implement OCR, image captioning, object detection, and a self-driving car agent. Finally, you'll move your NN model to production on the AWS Cloud. By the end of this book, you’ll be able to leverage modern NN architectures to solve over 50 real-world CV problems confidently.
Table of Contents (25 chapters)
1
Section 1 - Fundamentals of Deep Learning for Computer Vision
5
Section 2 - Object Classification and Detection
13
Section 3 - Image Manipulation
17
Section 4 - Combining Computer Vision with Other Techniques

Implementing deep Q-learning

So far, we have learned how to build a Q-table, which provides values that correspond to a given state-action combination by replaying a game – in this case, the Frozen Lake game – over multiple episodes. However, when the state spaces are continuous (such as a snapshot of a game of Pong), the number of possible state spaces becomes huge. We will address this in this section, as well as the ones to follow, using deep Q-learning. In this section, we will learn how to estimate the Q-value of a state-action combination without a Q-table by using a neural network hence the term deep Q-learning.

Compared to a Q-table, deep Q-learning leverages a neural network to map any given state-action (where the state can be continuous or discrete) combination to Q-values.

For this exercise, we will work on the CartPole environment in Gym. Here, our task is to balance the CartPole for as long as possible. The following image shows what the CartPole environment...