With Amazon Rekognition Custom Labels, you can easily build and deploy Machine Learning (ML) models to identify custom objects which are specific to your business domain in images without requiring advanced ML knowledge. When combined with Amazon Augmented AI (A2I), you can quickly integrate a ML workflow to capture and label images with a human workforce for model training. As ML lifecycle is an iterative and repetitive process, you need to implement an effective workflow that can provide for continuous model training with new data and automated deployment. Your workflow also needs to be flexible enough to allow for changes without requiring development rework as your business objectives change. Operationalizing an effective and flexible workflow can be resource intensive, especially for customers who have limited machine learning capabilities. In this post, we will use AWS Step Functions, AWS Lambda, and AWS System Manager Parameter Store to automate a configurable ML workflow for Rekognition Custom Labels and A2I. We will provide an overview of the solution and instructions to deploy it with AWS CloudFormation.
