Ultralytics v8.3.139 Release: Unified Data Export, Faster Setups, and More! 
Summary
Ultralytics v8.3.139 is here, bringing major upgrades to how you export, analyze, and share your YOLO validation metrics and prediction results! Easily export results in your preferred format for seamless sharing and integration, plus enjoy speedier installations and even smoother model deployment.
New Features
Unified Data Export
Export your metrics and prediction results in one go—with just a line of code—across all major task types:
- Detection
- Segmentation
- Classification
- Pose estimation
- Oriented bounding boxes
Supported export formats include DataFrame, CSV, XML, HTML, JSON, and direct SQLite database integration. This is enabled with the new DataExportMixin
class, making your workflow far more flexible and efficient!
PR by @RizwanMunawar: New DataExportMixin class for Metrics and Results exports
Improvements
-
Faster Dependency Management:
Install Python packages up to 3x faster by leveraging the “uv” installer when available for even smoother environments.
PR by @glenn-jocher: Use
uv
incheck_requirements
-
Enhanced Documentation:
Guides and docs now fully cover the new export options and clearly list all supported formats. -
Better Testing Coverage:
Added more robust test coverage to ensure all export features work reliably.
Bug Fixes
-
TFJS Export Compatibility:
Exporting models to TensorFlow.js now automatically disables grouped convolutions, preventing web deployment errors.
PR by @Y-T-G: Fix YOLO11 TFJS inference error
-
Raspberry Pi Test Optimization:
Slow similarity search tests are now skipped on Raspberry Pi, making testing faster and more reliable for embedded users.
PR by @RizwanMunawar: Disable similarity search tests on RaspberryPI
Why This Matters
- Simplifies Data Handling:
Choose the export format that suits your workflow—no more manual format conversions. - Boosts Productivity:
Instantly analyze results in Excel, database tools, or web dashboards, and share with teammates with ease. - Speeds Up Setup:
Spend less time waiting on package installs, thanks to the “uv” installer integration. - Streamlines Deployment:
Easier web model deployment and improved experience on Raspberry Pi devices. - Builds for the Future:
Sets the stage for even more flexible reporting and integrations everywhere Ultralytics is used.
How to Get Started
- Upgrade with:
pip install -U ultralytics
- Explore the new data export and deployment features in your next project!
- Check out the full changelog for all details.
- Release notes available on GitHub.
We’d love for you to try out Ultralytics v8.3.139 and share your feedback, questions, or suggestions. Your input drives our progress—let us know what you think and help shape the future of YOLO!
—
Special thanks to the Ultralytics contributors and community for pushing things forward!