Home/Blog/Cloud Computing/Using Amazon Rekognition for image and video analysis

Using Amazon Rekognition for image and video analysis

7 min read

Mar 06, 2025

content

Core capabilities of Amazon Rekognition

Image analysis

Video analysis

Use cases for Amazon Rekognition

PBS

Building a facial recognition system with Amazon Rekognition

Using Amazon Rekognition: Prerequisites

Benefits of using Rekognition

Final word

Imagine detecting faces, objects, and inappropriate content in images and videos—without building complex machine learning models. With Amazon Rekognition, businesses can integrate AI-powered image and video analysis in just a few API calls.

Amazon Rekognition is a fully managed AWS service that analyzes images and videos to identify objects, people, text, scenes, and activities. It also supports facial analysis, face comparison, and content moderation, making it a versatile tool for the security, media, and e-commerce industries.

Traditional computer vision solutions require specialized expertise, making them expensive and difficult to scale. Amazon Rekognition removes these barriers, allowing businesses to extract insights from visual data efficiently—without prior machine learning experience.

Core capabilities of Amazon Rekognition#

Amazon Rekognition offers robust image and video analysis capabilities, from face detection to content moderation. Its APIs simplify integrating these features into applications without requiring deep ML expertise.

Image analysis#

The image analysis technology of Amazon Rekognition is capable of providing multiple services. Some of them are as follows:

Object detection: Amazon Rekognition can identify and categorize objects, scenes, segments, or even concepts from images.
Content detection and filtering: Amazon Rekognition can filter inappropriate and unsafe text, content, and images. This allows for precise content moderation and enhanced safety.
Facial analysis: Amazon Rekognition offers powerful facial analysis services. It can not only detect and analyze faces, but can compare different faces in images and videos. It is also powerful enough to analyze emotions, sex, and age. This can help solve problems such as user verification and categorizing people according to their age or sex.

Video analysis#

The video analysis technology of Amazon Rekognition is capable of providing multiple services. Most of these services are similar to image analysis such as object, text, and unsafe content detection and facial analysis. The specific video analysis capabilities of Amazon Rekognition are as follows:

Identification in motion: Amazon Rekognition can track people’s movement across a video, mapping their path as they move through different frames—useful for security and surveillance applications. This capability allows Amazon Rekognition to create a trajectory of specific people in videos. It can also provide insights such as the individual’s location during tracking and facial landmarks in videos.
Video segmentation: Amazon Rekognition can identify useful and unimportant segments of videos. It can pinpoint the portions of videos that are not relevant to the topic of the content. This capability is useful for use cases such as video editing and indexing.
Face liveness: Face liveness detection is a security feature that verifies if a user is physically present during authentication. Amazon Rekognition achieves this by analyzing a short selfie video, detecting natural facial movements like eye blinks and head turns. It also identifies movement patterns to prevent spoofing attempts using photos or prerecorded videos, ensuring only real users can access resources.

Use cases for Amazon Rekognition#

Amazon Rekognition is a versatile AI service that can be applied across various industries to address challenges such as content moderation, user authentication, and media analysis. Some common use cases where Rekognition is used are given below:

Content moderation: Social media platforms can filter out offensive content using Amazon Rekognition to detect and flag inappropriate images and videos to maintain a safe environment.
Online examination: During an exam, Rekognition can monitor the video feed to detect any suspicious behavior or potential cheating, ensuring the integrity of the examination process.
Public safety: Law enforcement agencies can use Rekognition to analyze images and videos to identify victims and perpetrators, aiding in quicker resolutions.

Now that we have an idea of the common use cases of Amazon Rekognition let’s take a look at a couple of real-world examples where Amazon Rekognition is used:

Pinterest #

Pinterest is a visual discovery platform where users can upload and search for different media types, such as images, videos, and GIFs. It uses Amazon Rekognition for image and video analysis to filter out content that violates its policies. It also uses Rekognition to label the content uploaded by users and identify objects, scenes, and activities. This helps tag images with relevant keywords, improving the platform’s ability to recommend visually similar content.

PBS#

The American Public Broadcasting Service (PBS) uses Amazon Rekognition to manage its vast archives of video content. It identifies objects, faces, and activities in video frames and generates detailed metadata. This enables more precise search capabilities for internal teams and viewers to find relevant content efficiently.

PBS also uses Rekognition to detect inappropriate content in its archive to ensure its educational and family-friendly programming adheres to content guidelines.

Building a facial recognition system with Amazon Rekognition#

Now that we’ve explored Amazon Rekognition’s capabilities and real-world use cases let’s implement it. In this next section, we’ll walk through building a facial recognition system step by step using Rekognition APIs.

First, we’ll create a sign-up page where the user must upload their image. Once the image is uploaded, we’ll use the detect_faces function to verify whether it includes a picture of a human.

We’ll take the user to the login page once the user has signed up. Here, we’ll capture the user’s image using the built-in camera and use the compare_faces function to check if the image uploaded during sign-up and login belongs to the same person. If the similarity index between both images is greater than 75, we allow the user to log in to our system.

Let’s take a look at the components we’ll use to build this system:

S3 bucket: In this system, we’ll create an S3 bucket that will be used to store the user’s images.
Detect human: Create a function that uses the detect_faces feature in Amazon Rekognition to check whether the image upload has a human. This function will look like the following:

Note: In the code widgets below, we’ll use Boto3, the AWS SDK for Python.

import boto3
import sys
def check_for_human(bucket_name, object_key):
    # Initialize Rekognition client
    rekognition = boto3.client('rekognition')
    
    try:
        # Call Rekognition detect_faces function with S3 object
        response = rekognition.detect_faces(
            Image={
                'S3Object': {
                    'Bucket': bucket_name, #name of the S3 bucket
                    'Name': object_key #name of the image file in the S3 bucket
                }
            },
            Attributes=['ALL']  # Request additional detailed facial attributes
        )
        
        # Check if any faces were detected
        faces = response.get('FaceDetails', [])
        if not faces:
            raise ValueError("No human detected in the image.")
        elif len(faces) > 1:
            raise ValueError("Multiple faces detected. Please upload an image with only one face.")
        
        print(f"Detected 1 human face in the image. Confidence: {faces[0]['Confidence']:.2f}%")
    
    except Exception as e:
        print(f"Error: {str(e)}")

import boto3
def compare_faces_in_s3(bucket_name, image1, image2, similarity_threshold=75):
    try:
        # Create an Amazon Rekognition client
        rekognition = boto3.client(
            'rekognition'
        )
        # Compare the two images
        response = rekognition.compare_faces(
            SimilarityThreshold=similarity_threshold,
            SourceImage={'S3Object': {'Bucket': bucket_name, 'Name': image1}},
            TargetImage={'S3Object': {'Bucket': bucket_name, 'Name': image2}}
        )
        # Check for matches
        if len(response["FaceMatches"]) == 0:
            return 'Match not found'
        else:
            return f"Similarity: {response['FaceMatches'][0]['Similarity']:.2f}%"
    except Exception as e:
        return f"Error: {str(e)}"

Become an AWS Professional

Become an AWS Professional

Amazon Web Services (AWS) is the most comprehensive and widely used cloud platform in the world. AWS is used by millions of clients - including the fastest-growing startups, most prominent corporations, and top government agencies - to reduce costs, become more agile, and innovate faster. This path will lead you to develop web applications using AWS architecture and learn about linked entities like databases and networking. By the end, you'll have professional-level skills to develop AWS-based web applications confidently.

35hrs 45mins

Beginner

35 Playgrounds

47 Quizzes

Using Amazon Rekognition: Prerequisites#

Before diving into Amazon Rekognition, it’s important to understand the skills and tools required to get started with this service. For example, familiarity with cloud concepts, particularly AWS services like S3 for storing images and IAM for managing permissions, is essential.

Besides this, basic programming knowledge in a supported language (e.g., Python, JavaScript, or Java) is crucial to integrating Rekognition APIs. Additionally, ensure you have AWS credentials with the necessary permissions before building applications using Amazon Rekognition.

Benefits of using Rekognition#

Following are some benefits of using Amazon Rekognition for image and video analysis:

Scalability: As a fully managed service, Rekognition scales effortlessly to handle varying workloads, whether analyzing a few images or processing large video libraries.
Cost-effectiveness: Pay-as-you-go pricing ensures businesses only pay for what they use, making it suitable for organizations of all sizes.
High accuracy: Rekognition leverages AWS’s cutting-edge AI models to deliver precise results for various image and video analysis tasks.

Customizability: The custom labels feature lets businesses train Rekognition to detect specific objects, scenes, or concepts—without requiring ML expertise. This enables tailored solutions for tasks like brand logo recognition, defect detection in manufacturing, and medical image analysis.

Final word#

Amazon Rekognition offers robust image and video analysis tools, making it accessible for developers and businesses to integrate AI-driven insights into their applications. By leveraging its features, you can enhance security measures, automate content management, and deliver personalized user experiences.

Frequently Asked Questions

What image and video formats are supported by Amazon Rekognition?

Amazon Rekognition supports JPEG and PNG formats for images and MP4, MOV, and AVI formats encoded with H. 264 for videos.

What are Rekognition’s limits on image or video size?

For images, Amazon Rekognition can process files up to 15 MB when analyzed directly or up to 5 GB when stored in S3. There is no specific size limit for videos, but they must be stored in S3 for analysis.

How does Rekognition handle real-time video analysis?

Rekognition integrates with Amazon Kinesis Video Streams to perform real-time video analysis for surveillance or live monitoring activities.

Written By:

Fatima Mehmood

Join 2.5 million developers at

Explore the catalog

Free Resources

Languages supported by Amazon Rekognition
C++	GO	Java
Python (Boto3)	JavaScript	Kotlin
.Net	PHP	Ruby
Rust	Swift