Identify Product Defects using Industrial Computer Vision
Publication date: October 10, 2024 (Diagram history)
With this architecture, you can detect anomalies such as casting metal defects, damage,
and irregularities in X-ray images. You use Amazon Lookout for Vision
Product defects computer vision architecture diagram
The following steps describe the architecture:
-
Capture images under consistent conditions with X-ray machines, cameras, and other devices.
-
Transfer product images to AWS with AWS Transfer Family, AWS DataSync, or AWS IoT Greengrass for edge devices.
-
Store product images in Amazon S3 separated into train and test datasets.
-
Use Amazon Lookout for Vision on the training dataset to label, train, tune, and deploy the defect detection model.
-
Expose the model through Amazon API Gateway and Lambda for admins and data scientists to manage.
-
For runtime inference, use Lambda to start the model. Use AWS Step Functions with Lambda to orchestrate a serverless workflow.
-
For batch anomaly detection, submit a batch job to AWS Batch with AWS Fargate for compute.
-
Notify users on confidence level through Amazon Simple Notification Service (Amazon SNS). If the minimum confidence threshold is not met, the user provides input to label undetected data.
-
Store final results in Amazon S3 for Quick visualization. Write results back to the manufacturing execution system (MES) on the shop floor.
Further reading
For additional information, see the following resources:
Diagram history
To be notified about updates to this reference architecture diagram, subscribe to the RSS feed.
| Change | Description | Date |
|---|---|---|
Initial publication | Reference architecture diagram first published. | October 10, 2024 |
RSS subscription
To subscribe to RSS updates, you must have an RSS plugin enabled for the browser that you are using.
