Book Image

Modern Computer Vision with PyTorch

By : V Kishore Ayyadevara, Yeshwanth Reddy
5 (2)
Book Image

Modern Computer Vision with PyTorch

5 (2)
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

Implementing few-shot learning

Imagine a scenario where we give you only 10 images of a person and ask you to identify whether a new image is of the same person. As humans, we can classify such tasks with ease. However, the deep learning-based algorithms that we have learned so far would require hundreds/ thousands of labeled examples to classify accurately.

Multiple algorithms that fall in the meta-learning paradigm come in handy to solve this scenario. In this section, we will learn about Siamese networks, prototypical networks, and relation matching networks that work towards solving the few-images problem.

All three algorithms aim towards learning to compare two images to come up with a score for how similar the images are.

Here's an example of what to expect during few-shot classification:

In the preceding representative datasets, we have shown a few images of each class to the network while training and asked it to predict the class for a new image based on the images.