Upload training data to Amazon Lookout For Vision or import training data from Amazon Simple Storage Service (Amazon S3) and train a model. Admin users signup and login to a management front end website. Start or stop the model or do one-time defect detection by uploading an image. Camera or a client application invokes an Amazon API Gateway endpoint to get a signed URL from Amazon S3. The request is authorized by an AWS Lambda function and a signed URL is returned. Using the signed URL, an image, along with its associated metadata, is uploaded to Amazon S3. Image upload to the S3 bucket triggers an event notification to initiate an AWS Step Functions workflow. Fetch image from the S3 bucket and present to Amazon Lookout For Vision for anomaly detection using DetectAnomalies API. Store inference result and image metadata in Amazon DynamoDB. Publish a notification to an Amazon Simple Notification Service (Amazon SNS) topic to send an email alert to subscribed operators and plant managers in case a defect or a low-confidence result is detected. Inference results are fetched from DynamoDB Streams, transformed and enriched, sent to Amazon Kinesis Data Firehose for batching, and saved in another S3 bucket. Inference results datasets are imported into Amazon Quicksight. This process can be scheduled based on requirements. Create dashboards and analyses for business users, and gain insights from the inference results. Amazon CloudWatch provides a single pane of glass to operators and plant managers for workload and defect detection monitoring using logs, alarms, and dashboards. Alarm notifications from Amazon CloudWatch are sent to operators and plant managers by Amazon SNS whenever defects exceed a pre-defined thresh
