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 instance segmentation using Mask R-CNN

To help us understand how to code Mask R-CNN for instance segmentation, we will leverage a dataset that masks people who are present within an image. The dataset we'll be using has been created from a subset of the ADE20K dataset, which is available at https://groups.csail.mit.edu/vision/datasets/ADE20K/. We will only use those images where we have masks for people.

The strategy that we'll adopt is as follows:

  1. Fetch the dataset and then create datasets and dataloaders from it.
  2. Create a ground truth in a format needed for PyTorch's official implementation of Mask R-CNN.
  3. Download the pre-trained Faster R-CNN model and attach a Mask R-CNN head to it.
  4. Train the model with a PyTorch code snippet that has been standardized for training Mask R-CNN.
  5. Infer on an image by performing non-max suppression first and then identifying the bounding box and the mask corresponding to the people in the image.

Let's code up the preceding...