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

The concept of overfitting

So far, we've seen that the accuracy of the training dataset is typically more than 95%, while the accuracy of the validation dataset is ~89%.

Essentially, this indicates that the model does not generalize as much on unseen datasets since it can learn from the training dataset. This also indicates that the model is learning all the possible edge cases for the training dataset; these can't be applied to the validation dataset.

Having high accuracy on the training dataset and considerably lower accuracy on the validation dataset refers to the scenario of overfitting.
Some of the typical strategies that are employed to reduce the effect of overfitting are as follows:
  • Dropout
  • Regularization

We will look at what impact they have in the following sections.

Impact of adding dropout

We have already learned that whenever loss.backward() is calculated, a weight update happens. Typically, we would have hundreds of thousands of parameters within a network and...