Global Summit 2025 - AI Image Gallery Project

This repository contains three interconnected projects that together form an AI-powered image gallery application:

1. gallery-backend: Backend infrastructure and APIs
2. gallery-frontend: User interface and frontend application
3. image-generator: AI-based image generation utility

Project Overview

Architecture

Key Components:

• AWS Amplify: Provides frontend application hosting (using React).
• Amazon API Gateway + AWS Lambda: Used as backend API endpoints for image retrieval and upload.
• Amazon Bedrock: A managed service that utilizes foundation models through APIs. It uses the Amazon Nova Canvas model for image generation and the Claude 3.5 Sonnet v2 model for descriptions.
• Amazon SageMaker: Deploys the necessary model as an Endpoint to process image synthesis requests.
• Amazon Rekognition: Detects faces in images and videos, and crops the relevant facial areas.

Deploy

Requirements

• AWS Account with appropriate permissions
• Node.js v14 or higher
• Python 3.8 or higher
• AWS CDK installed and configured
• AWS CLI installed and configured

Gallery Backend

The backend infrastructure is built using AWS CDK and provides the foundation for the entire application. It consists of several key components:

• API Gateway: RESTful endpoints for image upload, retrieval, and user agreement management
• AWS Cognito: User authentication and authorization
• DynamoDB: Data storage for image processing status, display information, and user agreements
• Lambda Functions: Serverless processing for image manipulation and business logic
• S3 Buckets: Object storage for images and assets
• SageMaker Endpoints: AI model deployment for FaceChain and GFPGAN
• Container Infrastructure: ECR repositories and CodeBuild pipelines for AI model containers

The backend handles all data processing, storage, and AI model integration, providing a robust foundation for the application.

For deployment instructions, please refer to the gallery-backend README.

Gallery Frontend

The frontend is a React TypeScript application that provides the user interface for the gallery service. Key features include:

• User authentication with AWS Cognito
• Image upload and processing
• Face detection and manipulation
• AI-powered face swapping
• Multi-language support (i18n)
• User agreement management
• QR code generation for sharing

The frontend communicates with the backend APIs to provide a seamless user experience for image processing and management.

For deployment instructions, please refer to the gallery-frontend README.

Image Generator

The image generator is a Python-based utility that leverages Amazon Bedrock to create AI-generated images. It includes:

• Integration with Amazon Bedrock's Nova Canvas model for image generation
• Claude 3.5 Sonnet for text processing and prompt engineering
• Configuration for various image styles, historical periods, and attributes
• S3 integration for storing generated images
• DynamoDB integration for tracking image metadata

This component allows the application to generate high-quality base images that can be used in the face swapping process.

For deployment instructions, please refer to the image-generator README.

Amazon Bedrock Model Access Setup

Before using the image generator, you need to set up access to the required Amazon Bedrock models:

1. Access AWS Bedrock Console

1. Sign in to the AWS Bedrock console: https://console.aws.amazon.com/bedrock/
2. Set the region to us-east-1 (Nova Canvas model is only available in us-east-1)

2. Request Model Access

1. In the left navigation pane, under Bedrock configurations, choose Model access
2. Choose Modify model access
3. Select the following models:
   • Anthropic Claude 3.5 Sonnet
   • Amazon Nova Canvas
4. Choose Next
5. For Anthropic models, you must submit use case details
6. Review and accept the terms, then choose Submit

3. IAM Permissions Setup

To request model access, your IAM role needs the following permissions:

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": [
                "aws-marketplace:Subscribe",
                "aws-marketplace:Unsubscribe",
                "aws-marketplace:ViewSubscriptions"
            ],
            "Resource": "*"
        }
    ]
}

4. S3 Bucket Access Setup

To allow the image generator to save images to S3, you need to configure proper IAM permissions:

1. Go to AWS IAM Console: https://console.aws.amazon.com/iam/
2. Create a new IAM policy with the following permissions:

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": [
                "s3:PutObject",
                "s3:GetObject",
                "s3:ListBucket"
            ],
            "Resource": [
                "arn:aws:s3:::amazon-bedrock-gallery-sydney-jungseob",
                "arn:aws:s3:::amazon-bedrock-gallery-sydney-jungseob/*"
            ]
        }
    ]
}

3. Attach this policy to the IAM role/user that will be used to run the image generator
4. Make sure the S3 bucket name in the policy matches your actual bucket name

Important Notes

• Nova Canvas model is only available in us-east-1 region
• Model access approval may take several minutes
• If model access is denied, contact AWS support or choose alternative models

Deployment Instructions

Follow these steps to deploy the complete application:

1. Deploy Gallery Backend

First, deploy the backend infrastructure:

cd gallery-backend
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

# Generate unique ID and update cdk.context.json
UNIQUE_ID=$(openssl rand -hex 4)
sed -i '' "s/<your-unique-id>/$UNIQUE_ID/g" cdk.context.json

cdk synth
cdk deploy --all

This will create all necessary AWS resources including API Gateway, Lambda functions, DynamoDB tables, S3 buckets, and SageMaker endpoints.

2. Generate Base Images

After the backend is deployed, use the image generator to create base images:

cd ../image-generator
pip install -r requirements.txt
python generate_image.py

This will generate images using Amazon Bedrock and store them in the configured S3 bucket, making them available for the face swapping process.

3. Deploy Gallery Frontend

Finally, deploy the frontend application:

cd ../gallery-frontend
npm install
# Update .env file with your backend API endpoints
npm run build:prod
# Deploy the build directory to your hosting service of choice

Testing the Application

Once all components are deployed, you can test the complete workflow:

1. Access the frontend application
2. Create an account and sign in
3. Upload a photo for face swapping
4. Select a base image generated by the image generator
5. Process the image using the AI models deployed in the backend
6. View and share your generated images

Contributing

Please refer to the individual project READMEs for specific contribution guidelines.

Contacts

• Chulwoo Choi
• Jinwoo Park
• Jungseob Shin
• Seongjin Ahn
• Kihoon Kwon
• Jisoo Min
• Hyeryeong Joo

Contributors

License

The code of this projects is released under the MIT License. See the LICENSE file.

This software utilizes the pre-trained models. Users of this software must strictly adhere to these conditions of use.

Please note that if you intend to use this software for any commercial purposes, you will need to train your own models or find models that can be used commercially.