Amazon Rekognition

Hands-On

Demo

Agenda

In this demo, we will:

Explore Amazon Rekognition's different features
Create an Automation using Lambda, S3 and S3 Event Trigger
Create an S3 bucket to store images
Configure IAM permissions for Rekognition
Upload sample images for analysis
Perform Label detection
Test the complete setup
Clean up resources

Amazon Rekognition

Label detection

Image properties

Facial analysis

Text in image

Personal Protective Equipment (PPE) detection

Celebrity recognition

Build Automation Using S3, S3 Event Trigger, Lambda Function

Create bucket

rekognition-lab-demo-590121

Select trusted entity

AWSLambdaBasicExecutionRole

Add permissions

Role name

RekognitionLabLambdaRole

Review

Add some more permissions

AmazonS3FullAccess

Add some more permissions

AmazonRekognitionFullAccess

3 Permissions policies

Create Lambda Function

RekognitionImageAnalyzer

Change default execution role

Additional configurations

Lambda Code

import json
import boto3
import urllib.parse

# Initialize the Rekognition client
rekognition_client = boto3.client('rekognition')

def lambda_handler(event, context):
    print("Received event:", json.dumps(event, indent=2))

    # Iterate over each record in the event. S3 events are batched in a 'Records' list.
    for record in event.get('Records', []):
        # Get the bucket name and object key from the S3 event notification
        try:
            bucket_name = record['s3']['bucket']['name']
            # URL decode the object key in case of special characters (e.g., spaces)
            object_key = urllib.parse.unquote_plus(record['s3']['object']['key'])
        except KeyError as e:
            print(f"Error extracting bucket/key from S3 record: {e}")
            print("Skipping this record:", json.dumps(record))
            continue # Move to the next record

        print(f"Analyzing image: s3://{bucket_name}/{object_key}")

        try:
            # Call Rekognition DetectLabels API
            response = rekognition_client.detect_labels(
                Image={
                    'S3Object': {
                        'Bucket': bucket_name,
                        'Name': object_key
                    }
                },
                MaxLabels=10,        # Return up to 10 labels
                MinConfidence=75     # Return labels with at least 75% confidence
            )

            labels = response.get('Labels', [])

            print(f"Detected labels for {object_key}:")
            if labels:
                for label in labels:
                    print(f"- Label: {label['Name']}, Confidence: {label['Confidence']:.2f}%")
            else:
                print("No labels detected with the specified confidence.")

        except Exception as e:
            print(f"Error calling Rekognition or processing image '{object_key}': {e}")
            # For production, consider more specific error handling like for 'InvalidS3ObjectException'.
            # Depending on requirements, you might want to continue processing other records instead of returning an error.
            return {
                'statusCode': 500,
                'body': json.dumps(f'Error processing image: {str(e)}')
            }

    # Return a success response after processing all records
    return {
        'statusCode': 200,
        'body': json.dumps('All records processed successfully!')
    }