Demo of Rekognition recognizing an image

In this lesson we’ll demonstrate Amazon Rekognition’s core image-analysis capabilities and show how to use them from the AWS Console and programmatically via the AWS SDKs. Rekognition provides pre-built APIs to detect objects and scenes, analyze faces, detect text, moderate content, and more — useful for automated image tagging, content moderation, visual search, and identity workflows.

What Rekognition provides

Pre-trained image-analysis APIs with confidence scores and instance-level data (bounding boxes, landmarks).
Support for images stored in Amazon S3 or provided as raw image bytes.
Quick exploration in the AWS Console and production integration via SDKs (for example, boto3 for Python or the AWS SDK for JavaScript).

Useful links:

Rekognition documentation: https://docs.aws.amazon.com/rekognition/
boto3 Rekognition client: https://boto3.amazonaws.com/v1/documentation/api/latest/reference/services/rekognition.html
AWS SDKs & Tools: https://aws.amazon.com/tools/

Key features and typical APIs

Feature	API	Primary outputs	Common use cases
Label detection	`DetectLabels`	`Name`, `Confidence`, `Instances` (bounding boxes)	Auto-tagging, visual search, inventory detection
Image properties	(Console view) / part of image analysis	Dominant colors, color distribution, simple quality metrics	Theming, color-based filtering, cataloging
Content moderation	`DetectModerationLabels`	Moderation categories (e.g., Explicit Nudity, Violence) + confidence	User-upload screening, policy enforcement
Face detection & analysis	`DetectFaces`	Face bounding boxes, landmarks, attributes (age range, emotions, glasses, smile)	Cropping, UX overlays, emotion analysis (non-identifying use)
Face comparison	`CompareFaces`	Similarity scores, matched bounding boxes	Verification, authentication flows
Celebrity recognition	`RecognizeCelebrities`	Celebrity name, confidence, metadata	Media tagging, editorial workflows
Text detection (OCR)	`DetectText`	Detected text strings, confidence, bounding boxes	Signage reading, document data extraction

Label detection

Label detection identifies objects, scenes, and concepts inside images. Rekognition returns a list of labels with confidence scores and—when available—instance-level bounding boxes and per-instance confidence. Typical fields returned by DetectLabels: Name, Confidence, and Instances. Labels can include hierarchical or related concepts (for example, “Vehicle” and “Car”). Use cases: automatic tagging and search indexing, extracting structured metadata from image collections, or detecting inventory in photos. Here is an example of Rekognition detecting multiple objects, people, and a zebra crossing in a street photo. The console shows detected labels, confidence scores, and bounding boxes for each instance.

Image properties

The Image Properties view extracts visual attributes such as dominant colors, color distribution, and simple image-quality metrics. These attributes are useful for theme extraction, color search filters, or generating thumbnails consistent with brand palettes. The console shows primary colors detected in the image and foreground/background hints.

Image moderation

Rekognition’s moderation capability (DetectModerationLabels) flags potentially inappropriate or unsafe content. Labels include categories such as Explicit Nudity, Suggestive, and Violence, each with a confidence score. Common applications:

Automatic screening of user-uploaded images before publishing.
Enforcing content policies at scale.
Triaging content for human review.

Face analysis

Facial analysis detects faces and returns attributes and landmarks. Attributes include estimated age range, gender, emotions (HAPPY, SAD, CALM, etc.), whether the person is smiling, whether eyes/mouth are open, and presence of eyeglasses or sunglasses. Rekognition returns a confidence score for face detection and for many attributes, plus face landmarks (eyes, nose, mouth corners) and a bounding box for each face. Below is an example where Rekognition detects a face and reports attributes such as age range, emotions, and whether the subject is smiling.

A screenshot of the Amazon Rekognition facial analysis demo showing a red‑haired woman standing among shrubs with a blue box around her face. The right panel lists detected attributes (age range, emotions, glasses, eyes/mouth status, etc.).

Face comparison

CompareFaces compares a source face (for example, a selfie) with one or more target faces and returns similarity scores for any matches along with the bounding boxes of matched faces. This is useful for verification workflows, such as confirming a user’s identity against a stored image.

Celebrity recognition

RecognizeCelebrities can identify public figures. When Rekognition finds a match it returns the celebrity’s name, confidence score, and related metadata—useful for media tagging and editorial workflows.

Text detection (OCR)

DetectText extracts printed and many handwritten text strings from images, returning the detected text, a confidence score, and a bounding box for each line or word. Common uses:

Reading signage and storefront text.
Extracting fields from posters or product labels.
Automating data entry from scanned images.

Programmatic usage (example: Python + boto3)

For production you’ll typically call Rekognition via an AWS SDK. Rekognition supports images stored in S3 or supplied as raw image bytes. Below is an example using boto3 to call DetectLabels on an image in S3:

import boto3

client = boto3.client('rekognition', region_name='us-east-1')

response = client.detect_labels(
    Image={
        'S3Object': {
            'Bucket': 'my-bucket',
            'Name': 'images/zebra-crossing.jpg'
        }
    },
    MaxLabels=50,
    MinConfidence=75
)

for label in response['Labels']:
    print(f"{label['Name']}: {label['Confidence']:.1f}%")
    if 'Instances' in label:
        for inst in label['Instances']:
            print(f"  - Instance confidence: {inst.get('Confidence', 0):.1f}% bounding box: {inst.get('BoundingBox')}")

Best practices:

Prefer storing images in S3 for higher throughput and simpler permission management; grant Rekognition access via an IAM role.
For low-latency or transient workloads, you can pass image bytes directly to the API.

When using Rekognition in applications, prefer images stored in S3 for higher throughput and simpler permission management (grant Rekognition access via an IAM role). For real-time or low-latency cases you can send image bytes directly.

Privacy, compliance, and governance

Face recognition and identity-related features carry legal and ethical responsibilities. Depending on your jurisdiction and use case, you may need to:

Obtain user consent.
Document lawful basis and retention policies.
Provide user notices and opt-outs.
Follow AWS service guidelines and any applicable local laws.

Be mindful of privacy, local laws, and AWS terms when using face-based features. Avoid using facial recognition for sensitive decisions without appropriate governance and consent.

Summary

Amazon Rekognition offers a comprehensive set of image-analysis features—label detection, image properties, content moderation, face detection & analysis, face comparison, celebrity recognition, and OCR. Explore them in the AWS Console for quick validation, then integrate via the AWS SDKs to automate image analysis at scale. References and further reading:

Rekognition documentation: https://docs.aws.amazon.com/rekognition/
boto3 Rekognition client: https://boto3.amazonaws.com/v1/documentation/api/latest/reference/services/rekognition.html
AWS Identity and Access Management: https://docs.aws.amazon.com/iam/

Introduction

Services - Networking

Services - Storage

Services - Compute

Services - Database

Services - Application Integration

Services - Data and ML

Services - Migration and Transfer

Services - Management and Governance

Services - Security

Bringing it all together

Designing for Security

Designing for Reliability

Designing for Performance

Designing for Cost-Optimization

Applying your Design Skills

Demo of Rekognition recognizing an image

What Rekognition provides

Key features and typical APIs

Label detection

Image properties

Image moderation

Face analysis

Face comparison

Celebrity recognition

Text detection (OCR)

Programmatic usage (example: Python + boto3)

Privacy, compliance, and governance

Summary

Watch Video

​What Rekognition provides

​Key features and typical APIs

​Label detection

​Image properties

​Image moderation

​Face analysis

​Face comparison

​Celebrity recognition

​Text detection (OCR)

​Programmatic usage (example: Python + boto3)

​Privacy, compliance, and governance

​Summary

Watch Video

What Rekognition provides

Key features and typical APIs

Label detection

Image properties

Image moderation

Face analysis

Face comparison

Celebrity recognition

Text detection (OCR)

Programmatic usage (example: Python + boto3)

Privacy, compliance, and governance

Summary