This code sample illustrates a minimal User Generated Content Moderation Pipeline for video files. It makes use of Rekognition Moderation capabilities to scan video content for inappropriate images. The rationale behind this code sample is to block UGC from progressing further in the analysis pipeline, so that customers don't perform AI/ML workloads on inappropriate content.

• Users upload media content to the inspection bucket deployed by the code sample
• An upload notification is sent on EventBridge which triggers the UGC state machine
• if the video is clear from inappropriate content, it gets replicated to a destination bucket
• otherwise, an email is sent to the moderation team for further inspection

REPLICATION_BUCKET_NAME=<YOUR_BUCKET_NAME> # for example "my-bucket-name"
NOTIFICATION_EMAIL_ADDRESS=<YOUR_EMAIL_ADDRESS> # for example "you@yourcompany.com"

npm i
cd app && npm i && cd ..

cdk deploy \
    --parameters replicationBucketName=$REPLICATION_BUCKET_NAME \
    --parameters notificationEmailAddress=$NOTIFICATION_EMAIL_ADDRESS \
    --outputs-file assets/cdk-outputs.json

replicationBucketName is the bucket where the object will be replicated if it does not contain inappropriate content. This must be in the same region as this stack.
notificationEmailAddress is valid email address where you'll receive moderation events notifications. Expect a confirmation email to this address from AWS SNS. Make sure you confirm the SNS subscription.

You can start the workflow by uploading a video to your S3 bucket at the /public/upload/ path. For example, run the following

INSPECTION_BUCKET=$(cat ./assets/cdk-outputs.json | jq .BasupUgcStackStack.InspectionBucketName | xargs)
UPLOAD_FILE_PATH=path/to/your/video

aws s3 cp $UPLOAD_FILE_PATH s3://$INSPECTION_BUCKET/public/upload/

You can monitor the execution of the flow in your StepFunction Console

This code sample has been developed by the UKIR Green Field Startup Solutions Architecture Team @AWS.