Computer Vision on AWS Build and Deploy Real-World CV Solutions with Amazon Rekognition, Lookout for Vision, and SageMaker
Develop scalable computer vision solutions for real-world business problems and discover scaling, cost reduction, security, and bias mitigation best practices with AWS AI/ML services Purchase of the print or Kindle book includes a free PDF eBook Key Features Learn how to quickly deploy and automate...
| Autor principal: | |
|---|---|
| Otros Autores: | , |
| Formato: | Libro electrónico |
| Idioma: | Inglés |
| Publicado: |
Birmingham :
Packt Publishing, Limited
2023.
|
| Edición: | 1st ed |
| Materias: | |
| Ver en Biblioteca Universitat Ramon Llull: | https://discovery.url.edu/permalink/34CSUC_URL/1im36ta/alma991009731836506719 |
Tabla de Contenidos:
- Cover
- Title Page
- Copyright and Credits
- Contributors
- Table of Contents
- Preface
- Part 1: Introduction to CV on AWS and Amazon Rekognition
- Chapter 1: Computer Vision Applications and AWS AI/ML Services Overview
- Technical requirements
- Understanding CV
- CV architecture and applications
- Data processing and feature engineering
- Data labeling
- Solving business challenges with CV
- Contactless check-in and checkout
- Video analysis
- Content moderation
- CV at the edge
- Exploring AWS AI/ML services
- AWS AI services
- Amazon SageMaker
- Setting up your AWS environment
- Creating an Amazon SageMaker Jupyter notebook instance
- Summary
- Chapter 2: Interacting with Amazon Rekognition
- Technical requirements
- The Amazon Rekognition console
- Using the Label detection demo
- Examining the API request
- Examining the API response
- Other demos
- Monitoring Amazon Rekognition
- Quick recap
- Detecting Labels using the API
- Uploading the images to S3
- Initializing the boto3 client
- Detect the Labels
- Using the Label information
- Using bounding boxes
- Quick recap
- Cleanup
- Summary
- Chapter 3: Creating Custom Models with Amazon Rekognition Custom Labels
- Technical requirements
- Introducing Amazon Rekognition Custom Labels
- Benefits of Amazon Rekognition Custom Labels
- Creating a model using Rekognition Custom Labels
- Deciding the model type based on your business goal
- Creating a model
- Improving the model
- Starting your model
- Analyzing an image
- Stopping your model
- Building a model to identify Packt's logo
- Step 1 - Collecting your images
- Step 2 - Creating a project
- Step 3 - Creating training and test datasets
- Step 4 - Adding labels to the project
- Step 5 - Drawing bounding boxes on your training and test datasets
- Step 6 - Training your model.
- Validating that the model works
- Step 1 - Starting your model
- Step 2 - Analyzing an image with your model
- Step 3 - Stopping your model
- Summary
- Part 2: Applying CV to Real-World Use Cases
- Chapter 4: Using Identity Verification to Build a Contactless Hotel Check-In System
- Technical requirements
- Prerequisites
- Creating the image bucket
- Uploading the sample images
- Creating the profile table
- Introducing collections
- Creating a collection
- Describing a collection
- Deleting a collection
- Quick recap
- Describing the user journeys
- Registering a new user
- Authenticating a user
- Registering a new user with an ID card
- Updating the user profile
- Implementing the solution
- Checking image quality
- Indexing face information
- Search existing faces
- Quick recap
- Supporting ID cards
- Reading an ID card
- Using the CompareFaces API
- Quick recap
- Guidance for identity verification on AWS
- Solution overview
- Deployment process
- Cleanup
- Summary
- Chapter 5: Automating a Video Analysis Pipeline
- Technical requirements
- Creating the video bucket
- Uploading content to Amazon S3
- Creating the person-tracking topic
- Subscribing a message queue to the person-tracking topic
- Creating the person-tracking publishing role
- Setting up IP cameras
- Quick recap
- Using IP cameras
- Installing OpenCV
- Installing additional modules
- Connecting with OpenCV
- Viewing the frame
- Uploading the frame
- Reporting frame metrics
- Quick recap
- Using the PersonTracking API
- Uploading the video to Amazon S3
- Using the StartPersonTracking API
- Receiving the completion notification
- Using the GetPersonTracking API
- Reviewing the GetPersonTracking response
- Viewing the frame
- Quick recap
- Summary
- Chapter 6: Moderating Content with AWS AI Services
- Technical requirements.
- Moderating images
- Using the DetectModerationLabels API
- Using top-level categories
- Using secondary-level categories
- Putting it together
- Quick recap
- Moderating videos
- Creating the supporting resources
- Finding the resource ARNs
- Uploading the sample video to Amazon S3
- Using the StartContentModeration API
- Examining the completion notification
- Using the GetContentModeration API
- Quick recap
- Using AWS Lambda to automate the workflow
- Implement the Start Analysis Handler
- Implementing the Get Results Handler
- Publishing function changes
- Experiment with the end-to-end
- Summary
- Part 3: CV at the edge
- Chapter 7: Introducing Amazon Lookout for Vision
- Technical requirements
- Introducing Amazon Lookout for Vision
- The benefits of Amazon Lookout for Vision
- Creating a model using Amazon Lookout for Vision
- Choosing the model type based on your business goals
- Creating a model
- Starting your model
- Analyzing an image
- Stopping your model
- Building a model to identify damaged pills
- Step 1 - collecting your images
- Step 2 - creating a project
- Step 3 - creating the training and test datasets
- Step 4 - verifying the dataset
- Step 5 - training your model
- Validating it works
- Step 1 - trial detection
- Step 2 - starting your model
- Step 3 - analyzing an image with your model
- Step 4 - stopping your model
- Summary
- Chapter 8: Detecting Manufacturing Defects Using CV at the Edge
- Technical requirements
- Understanding ML at the edge
- Deploying a model at the edge using Lookout for Vision and AWS IoT Greengrass
- Step 1 - Launch an Amazon EC2 instance
- Step 2 - Create an IAM role and attach it to an EC2 instance
- Step 3 - Install AWS IoT Greengrass V2
- Step 4 - Upload training and test datasets to S3
- Step 5 - Create a project.
- Step 6 - Create training and test datasets
- Step 7 - Train the model
- Step 8 - Package the model
- Step 9 - Configure IoT Greengrass IAM permissions
- Step 10 - Deploy the model
- Step 11 - Run inference on the model
- Step 12 - Clean up resources
- Summary
- Part 4: Building CV Solutions with Amazon SageMaker
- Chapter 9: Labeling Data with Amazon SageMaker Ground Truth
- Technical requirements
- Introducing Amazon SageMaker Ground Truth
- Benefits of Amazon SageMaker Ground Truth
- Automated data labeling
- Labeling Packt logos in images using Amazon SageMaker Ground Truth
- Step 1 - collect your images
- Step 2 - create a labeling job
- Step 3 - specify the job details
- Step 4 - specify worker details
- Step 5 - providing labeling instructions
- Step 6 - start labeling
- Step 7 - output data
- Importing the labeled data with Rekognition Custom Labels
- Step 1 - create the project
- Step 2 - create training and test datasets
- Step 3 - model training
- Summary
- Chapter 10: Using Amazon SageMaker for Computer Vision
- Technical requirements
- Fetching the LabelMe-12 dataset
- Installing TensorFlow 2.0
- Installing matplotlib
- Using the built-in image classifier
- Upload the dataset to Amazon S3
- Prepare the job channels
- Start the training job
- Monitoring and troubleshooting
- Quick recap
- Handling binary metadata files
- Declaring the Label class
- Reading the annotations file
- Declaring the Annotation class
- Validate parsing the file
- Restructure the files
- Load the dataset
- Quick recap
- Summary
- Part 5: Best Practices for Production-Ready CV Workloads
- Chapter 11: Integrating Human-in-the-Loop with Amazon Augmented AI (A2I)
- Technical requirements
- Introducing Amazon A2I
- Core concepts of Amazon A2I
- Learning how to build a human review workflow.
- Creating a labeling workforce
- Setting up an A2I human review workflow or flow definition
- Initiating a human loop
- Leveraging Amazon A2I with Amazon Rekognition to review images
- Step 1 - Collecting your images
- Step 2 - Creating a work team
- Step 3 - Creating a human review workflow
- Step 4 - Starting a human loop
- Step 5 - Checking the human loop status
- Step 6 - Reviewing the output data
- Summary
- Chapter 12: Best Practices for Designing an End-to-End CV Pipeline
- Defining a problem that CV can solve and processing data
- Developing a CV model
- Training
- Evaluating
- Tuning
- Deploying and monitoring a CV model
- Shadow testing
- A/B testing
- Blue/Green deployment strategy
- Monitoring
- Developing an MLOps strategy
- SageMaker MLOps features
- Workflow automation tools
- Using the AWS Well-Architected Framework
- Cost optimization
- Operational excellence
- Reliability
- Performance efficiency
- Security
- Sustainability
- Summary
- Chapter 13: Applying AI Governance in CV
- Understanding AI governance
- Defining risks, documentation, and compliance
- Data risks and detecting bias
- Auditing, traceability, and versioning
- Monitoring and visibility
- MLOps
- Responsibilities of business stakeholders
- Applying AI governance in CV
- Types of biases
- Mitigating bias in identity verification workflows
- Using Amazon SageMaker for governance
- ML governance capabilities with Amazon SageMaker
- Amazon SageMaker Clarify for explainable AI
- Summary
- Index
- Other Books You May Enjoy.
