#Quickstart

#S3 Integration

Label Studio supports integration with S3-compatible storage for importing data and exporting annotations. Common options include Amazon S3, Ceph RGW, and other S3-compatible object storage services.

#Prerequisites

• S3-compatible storage bucket with appropriate permissions
• Access credentials (Access Key ID and Secret Access Key)

The bucket and credentials must come from a supported S3-compatible service; no specific object storage product is required by the platform.

#Using S3 with Label Studio

1. Access Storage Settings

   • Open Label Studio project
   • Go to Settings > Cloud Storage

2. Add Source Storage

   • Click Add Source Storage
   • Select AWS S3 as storage type
   • Fill in the required information:
     • Storage Title: Name for the storage connection
     • Bucket Name: S3 bucket name
     • Region Name: Storage region (e.g., us-east-1 for AWS S3; may be empty if the provider does not require it)
     • S3 Endpoint: Optional custom S3 endpoint (leave empty for AWS S3 default; set when using a non-default or S3-compatible endpoint)
     • Access Key ID: Access key
     • Secret Access Key: Secret key
     • Session Token: Optional session token for temporary credentials
     • Bucket Prefix: Optional path prefix in the bucket (e.g., data/, input/)
     • File Filter Regex: Optional regex to filter files (e.g., .*csv or .*(jpe?g|png|tiff))
   • Configure optional settings:
     • Treat every bucket object as a source file: Check for media files, uncheck for JSON task files
     • Recursive scan: Enable to scan subdirectories recursively
     • Use pre-signed URLs: Enable for direct browser access to S3 (recommended)
     • Expiration minutes: URL expiration time (default: 15 minutes) when Use pre-signed URLs enabled
   • Click Check Connection to test connectivity
   • Click Add Storage to create the storage connection

3. Add Target Storage (Optional)

   • Click Add Target Storage to export annotations to S3
   • Fill in similar S3 parameters like Source Storage
   • Additional Target Storage parameters:
     • SSE KMS Key ID: Optional KMS key for server-side encryption
   • Configure optional settings:
     • Can delete objects from storage: Enable to allow deletion of annotations from storage
   • Click Check Connection to test connectivity
   • Click Add Storage to create the storage connection

4. Upload Data to S3

   • Upload data files to the configured S3 bucket and prefix path
   • Ensure data files are accessible with the configured access credentials
   • Use the AWS CLI or another S3-compatible client for bulk uploads

5. Import Data

   • Click Sync Storage under Source Cloud Storage to import data from S3
   • Use sync whenever new data is added to the S3 bucket

6. Perform Annotations

   • Access the imported data in Label Studio interface
   • Complete annotations using the configured labeling interface

7. Export Annotations

   • Click Export button to download annotation results in various formats (JSON, CSV, etc.)
   • Or click Sync Storage for Target Cloud Storage to push annotations to S3
   • Note: Target Storage exports annotations in JSON format only. Use Label Studio SDK to convert JSON annotations to other formats (CSV, COCO, Pascal VOC, YOLO, etc.). See SDK converter for details.

8. Apply Data and Annotations to Model Training/Validation

   • Download training data and annotations from S3 using the AWS CLI, boto3, or another S3 client. See S3 examples for implementation details.
   • Convert annotation format using Label Studio SDK if needed.
   • Integrate data into machine learning pipelines.
   • Use annotations for model training or validation.

#Storage Structure Suggestions

• Use different buckets or different path prefixes for different projects to avoid data conflicts.
• Target and Source can use the same S3 bucket with different path prefixes (e.g., input/ for source, output/ for target), or use different buckets for better data isolation and access control.

#Additional Resources

For Label Studio quickstart guide, please refer to the official documentation: Getting Started With Label Studio: A Step-By-Step Guide