New Release: Ultralytics v8.3.91

Discussion
,
1

:rocket: Announcing Ultralytics YOLO v8.3.91 Release

:glowing_star: Summary

We’re thrilled to announce the release of YOLO v8.3.91! This update focuses on simplifying TensorFlow installation, advancing export compatibility for ARM64/Linux platforms, improving dataset handling, and refining both documentation and visualization features. A smoother, smarter experience awaits you—let’s dive into the highlights!

:bar_chart: Key Changes

TensorFlow Installation Simplification

Simplified the TensorFlow setup for a seamless export experience. Dependency requirements have been updated to minimize hurdles in the installation process.

Export Enhancements

  • Platform Support: Enhancements in TFLite and TensorFlow.js export ensure smoother compatibility with ARM64/Linux environments.
  • Error Handling: Clearer, actionable error messages are now integrated for unsupported configurations during export.

Dataset Handling Improvements

  • Automatic Fallbacks: Missing val or test dataset splits are now automatically handled with fallback logic to streamline training workflows.
  • Enhanced Logging: Improved batch-by-batch logging of image data during training makes integration with Comet more effective.

Visualization Updates

Annotation readability has been prioritized with optimized font sizes, making debugging and collaboration smoother during the training process.

Documentation Updates

  • Compared performance metrics of YOLO models, such as YOLO11n-seg, against Meta’s SAM models to illustrate YOLO’s efficiency and versatility.
  • Added social icons for broader engagement, including WeChat, to expand accessibility and support for our community.

:bullseye: Purpose & Impact

This release is all about making the Ultralytics YOLO ecosystem more efficient, versatile, and user-friendly:

  • Enhanced Compatibility: Exports are now robust across ARM64/Linux platforms, bridging gaps for developers working on diverse hardware setups.
  • Improved User Experience: Simplified TensorFlow installation and automated dataset handling reduce errors and save you precious time.
  • Streamlined Workflows: Updated visualizations and onboard logging refine the training pipeline, supporting clearer debugging and smoother experimentation.
  • Accessible Resources: New benchmarking comparisons and expanded social media support ensure that educational and practical resources are more readily available.

With v8.3.91, you’re equipped for smarter development, enhanced collaboration, and maximum performance across your YOLO-based projects!

:counterclockwise_arrows_button: What’s Changed

Take a closer look at the key contributions driving this release:

  • Improved Visualization: Enhanced label readability in classify tasks for better Comet integration (#19700 by @yaricom)
  • WeChat Social Icon: Added WeChat to Docs for extended community accessibility (#19702 by @glenn-jocher)
  • Benchmark Comparisons: Updated SAM vs YOLO documentation to guide model selection (#19705 by @glenn-jocher)
  • TensorFlow Installation: Streamlined TensorFlow dependencies (#19712 by @glenn-jocher)

For a complete list of changes, explore the Full Changelog.

:light_bulb: Try It & Share Feedback

We invite you to experience these improvements firsthand. Upgrade to the latest release with:

pip install ultralytics --upgrade

Explore the new features, put the updated export capabilities to the test, and share your feedback with us. Being a part of this community-driven journey helps shape the future of YOLO for everyone.

Release details and download options are available on the official release page.

Thank you for being an integral part of the YOLO ecosystem. Together, we’re pushing the boundaries of open-source innovation! :rocket: