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)
Section 1 - Fundamentals of Deep Learning for Computer Vision
Section 2 - Object Classification and Detection
Section 3 - Image Manipulation
Section 4 - Combining Computer Vision with Other Techniques

Understanding the impact of batch normalization

Previously, we learned that when the input value is large, the variation of the Sigmoid output doesn't make much difference when the weight values change considerably.

Now, let's consider the opposite scenario, where the input values are very small:

When the input value is very small, the Sigmoid output changes slightly, making a big change to the weight value.

Additionally, in the Scaling the input data section, we saw that large input values have a negative effect on training accuracy. This suggests that we can neither have very small nor very big values for our input.

Along with very small or very big values in input, we may also encounter a scenario where the value of one of the nodes in the hidden layer could result in either a very small number or a very large number, resulting in the same issue we saw previously with the weights connecting the hidden layer to the next layer.

Batch normalization comes to the rescue in such...