DEV Community

Cover image for Amazon Bedrock Image Analysis Tutorial with AWS Lambda
Pratik Ponde
Pratik Ponde

Posted on

Amazon Bedrock Image Analysis Tutorial with AWS Lambda

#ai #beginners #devops #design

👋 Hey there! This is Pratik, a Senior DevOps Consultant with a strong background in automating and optimizing cloud infrastructure, particularly on AWS. Over the years, I have designed and implemented scalable solutions for enterprises, focusing on infrastructure as code, CI/CD pipelines, cloud security, and resilience. My expertise lies in translating complex cloud requirements into efficient, reliable, and cost-effective architectures.

Looking Beyond OCR

Traditional OCR systems are great at extracting text but they fall short when the task shifts from reading content to recognizing visual patterns. What if your goal isn’t to read a document, but to determine whether a specific reference image appears within it regardless of where it’s placed, how it’s rotated, or how it’s scaled?

In this article, we’ll build an AI-powered system using AWS Lambda, S3, and Amazon Bedrock that can:

  • Upload a reference image
  • Upload a target document (Image)
  • Detect whether the reference image exists inside the target file
  • Return a true/false result with high accuracy

Architecture Diagram

Architecture Explanation

The system follows a simple serverless flow using Amazon S3, AWS Lambda, and Amazon Bedrock.

Step-by-Step Flow

1. User Uploads Files to S3

  • The user uploads:
    • A reference image (e.g., logo/signature)
    • A target file (image)
  • Both are stored in Amazon S3

2. S3 Triggers Lambda

  • Once the files are uploaded, an event notification triggers the Lambda function
  • Lambda acts as the orchestrator

3. Lambda Processes Data

  • Fetches:

    • Reference image
    • Target file

  • Converts them into base64 format

  • Prepares a structured prompt

4. Bedrock Performs Analysis

  • Lambda sends both images to Amazon Bedrock
  • A multimodal model analyzes:
    • Shape
    • Orientation
    • Position
    • Partial visibility

5. Result Returned

  • Bedrock responds with:
    • "true" → Image found
    • "false" → Image not found

The result will be stored in the S3 bucket.

Step-by-Step Implementation

1️⃣ Create S3 Bucket Structure

Go to Amazon S3 and create bucket with name “image-matching-bucket”

Inside it, create folders:

reference-images/
target-files/
Enter fullscreen mode Exit fullscreen mode

Upload:
• Reference image → reference-images/logo.png
• Target file → target-files/doc1.png

2️⃣ Create IAM Role for Lambda

Go to AWS Identity and Access Management and Create a role with:

Permissions:

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Action": [
                "s3:GetObject"
            ],
            "Effect": "Allow",
            "Resource": "arn:aws:s3:::iimage-matching-bucket/*"
        },
        {
            "Action": [
                "bedrock:InvokeModel",
                "aws-marketplace:Subscribe",
                "aws-marketplace:ViewSubscriptions"
            ],
            "Effect": "Allow",
            "Resource": "*"
        }
    ]
}
Enter fullscreen mode Exit fullscreen mode

3️⃣ Create Lambda Function

Go to AWS Lambda

  • Runtime: Node.js 18+
  • Attach IAM role created above

 

Lambda Code is available on GitHub Repository

🔗 GitHub Repository:

GitHub logo pratiksponde / Amazon-Bedrock-and-Lambda

Intelligent Image Presence Detection using AWS Bedrock

Detect whether a reference image exists inside a target document (image) using AI-powered visual reasoning with AWS services.

This project goes beyond traditional OCR by identifying images even if they are:

  • Rotated
  • Resized
  • Partially visible
  • Positioned anywhere

📖 Full article:
https://dev.to/pratik_26/amazon-bedrock-image-analysis-tutorial-with-aws-lambda-4mi9

🚀 Architecture

This solution uses:

  • Amazon S3 – Store reference and target files
  • AWS Lambda – Process and orchestrate logic
  • Amazon Bedrock – Perform multimodal AI analysis

🔄 Flow

User → Upload to S3 → S3 Event → Lambda → Bedrock → Result Store in S3 Bucket (true/false)

Features

  • Detect image presence inside documents
  • Supports images and PDFs
  • Works with rotated, scaled, or partial matches
  • Fully serverless and scalable
  • Uses multimodal AI (no traditional CV required)

Tech Stack

  • Node.js (AWS Lambda)
  • AWS SDK v3
  • Amazon Bedrock (Claude Vision / Multimodal Model)
  • Amazon S3

📂 Project Structure


├── lambda/
│   └── index.js
…
View on GitHub

Feel free to clone the repository and try it in your own AWS environment.

4️⃣ Configure S3 Trigger

In S3 → Properties → Event Notifications

Add event:

  • Event type: POST
  • Prefix: target-files/
  • Destination: Lambda

This ensures Lambda runs automatically when a file is uploaded.

5️⃣ Test the Flow

Upload:

  • Reference image → already uploaded
  • Upload new target file

Expected Result:

Lambda runs → Bedrock analyzes → returns → Stores in S3 Bucket

{
  "result": "true"
}
Enter fullscreen mode Exit fullscreen mode

Key Takeaways

  • You can go beyond OCR using multimodal AI
  • Serverless architecture makes it scalable and cost-efficient
  • Prompt engineering plays a critical role in accuracy
  • This approach is ideal when visual context matters more than text

Conclusion

By combining AWS Lambda, Amazon S3, and Amazon Bedrock, we built an intelligent system that can understand and compare images inside documents, something traditional OCR tools struggle with.

This is a step toward building AI-native document processing systems that can reason visually, not just read text.

Let’s Keep the Conversation Going

Have thoughts, questions, or any experience with Generative AI to share? I would love to hear from you! Feel free to leave a comment or connect with me on LinkedIn. Let's learn and grow together as a community of builders.
Keep exploring, keep automating and see you in the next one!

Top comments (0)