Book Image

Computer Vision on AWS

By : Lauren Mullennex, Nate Bachmeier, Jay Rao
Book Image

Computer Vision on AWS

By: Lauren Mullennex, Nate Bachmeier, Jay Rao

Overview of this book

Computer vision (CV) is a field of artificial intelligence that helps transform visual data into actionable insights to solve a wide range of business challenges. This book provides prescriptive guidance to anyone looking to learn how to approach CV problems for quickly building and deploying production-ready models. You’ll begin by exploring the applications of CV and the features of Amazon Rekognition and Amazon Lookout for Vision. The book will then walk you through real-world use cases such as identity verification, real-time video analysis, content moderation, and detecting manufacturing defects that’ll enable you to understand how to implement AWS AI/ML services. As you make progress, you'll also use Amazon SageMaker for data annotation, training, and deploying CV models. In the concluding chapters, you'll work with practical code examples, and discover best practices and design principles for scaling, reducing cost, improving the security posture, and mitigating bias of CV workloads. By the end of this AWS book, you'll be able to accelerate your business outcomes by building and implementing CV into your production environments with the help of AWS AI/ML services.
Table of Contents (21 chapters)
Part 1: Introduction to CV on AWS and Amazon Rekognition
Part 2: Applying CV to Real-World Use Cases
Part 3: CV at the edge
Part 4: Building CV Solutions with Amazon SageMaker
Part 5: Best Practices for Production-Ready CV Workloads

Handling binary metadata files

Ingesting the hotdog-nothotdog dataset was pretty straightforward since you only needed to parse the path. However, many scientific and open source datasets encode their labels into binary files. This approach reduces the data size and increases file parsing performance. You’re probably wondering whether we can still use them. Of course!

The LabelMe-12 dataset from the technical requirements section is one such example. It includes the label information in the annotation.bin (binary) and annotation.txt (human-readable) files under the ./data/test and ./data/train folders. Let’s focus on the binary file and only use the human-readable copy for troubleshooting.

We will perform the following steps to do this:

  1. Declare the custom Label enumeration.
  2. Declare the custom Annotation class.
  3. Read each image’s labels from the file.
  4. Confirm the expected counts are present.
  5. Normalize the file structure.