Today we’re introducing a new cloud storage option in CVAT: Backblaze B2, Backblaze’s S3‑compatible object storage service. 

This integration allows you to connect your annotation projects directly to Backblaze buckets, enabling scalable, cost-effective dataset storage within your existing ML pipelines.

Why connect CVAT to Backblaze B2?

Computer vision datasets scale rapidly into terabytes, making local disks and manual synchronization scripts inefficient. Backblaze B2 provides an affordable alternative to traditional cloud providers for hosting raw images, videos, and annotation files.

Key operational benefits include:

• Centralized Storage: Keep raw media and annotation files in the same bucket.
• Pipeline Automation: Eliminate manual exports by connecting CVAT directly to the storage tier used by your training and MLOps pipelines.
• API Compatibility: Leverage Backblaze's S3-compatible API to interact with datasets using existing SDKs and CLIs.

Supported Capabilities

Backblaze B2 functions as a native cloud storage provider inside the CVAT interface. Once configured, you can:

• Stream images and videos directly from a B2 bucket to create CVAT annotation tasks.
• Save annotations, backups, and derived artifacts back to the source bucket.
• Isolate data access by mapping different CVAT projects to specific B2 buckets or prefixes.

How to Get Started

To connect your storage:

1. Navigate to the Cloud Storages section in CVAT.
2. Add a new storage and select AWS S3 as the provider type.
3. Input your Backblaze B2 bucket name, endpoint URL, Access Key ID, and Secret Access Key.

For detailed configuration steps, permission policies, and credential management, read the Backblaze B2 Storage Guide.

Have feedback or feature requests for this integration? Let us know in our community channels.