New Release: Ultralytics v8.3.114

:rocket: Ultralytics v8.3.114 Release: Smarter Object Tracking with Native ReID, Docs & Export Enhancements!

:glowing_star: Summary

We’re excited to announce Ultralytics v8.3.114! This release brings robust tracker re-identification (ReID) for object tracking, key improvements in documentation, refined dataset handling, and greater reliability for model export. It’s all about smarter tracking, clearer guides, and smoother deployments. Dive in to experience a more powerful, user-friendly Ultralytics ecosystem! :mechanical_arm:


:tada: New Features

Tracker Re-Identification (ReID) Support

  • Native BoTSORT ReID Integration:
    Re-identify and reconnect lost objects with advanced feature-based and model-based ReID, now natively built into the BoTSORT tracker.
  • Automatic, Smarter Association:
    The tracker’s new default auto mode uses YOLO model features for ReID whenever possible—improving tracking in complex scenarios like occlusion and crowd scenes.
  • More Robust Performance:
    Enhanced extraction and association logic ensure tracks can reliably recover lost identities and maintain continuity.

Pull request: New tracker re-identification (ReID) of lost tracks by @Y-T-G


:books: Documentation Enhancements


:hammer_and_wrench: Improvements & Bug Fixes


:bullseye: Purpose & Impact

  • Smarter Object Tracking:
    The new ReID feature allows trackers to reconnect with lost tracks, boosting reliability in demanding use cases with minimal configuration.
  • More Accessible Documentation:
    Improvements help both beginners and experienced users to get started and troubleshoot issues faster.
  • Smoother Export & Deployment:
    Enhanced export routines lead to fewer errors and confusion when deploying models in various environments.

:loudspeaker: How to Try It

Ready to upgrade? You can update via pip:

pip install -U ultralytics

Explore the official Ultralytics v8.3.114 release notes and see the full changelog from v8.3.113…v8.3.114 for all the details.


:ear: We Value Your Feedback!

Your experience and suggestions continue to drive YOLO’s progress—please give this release a try and share your thoughts in this thread! Let us know what works well and where you see room for improvement. If you encounter any issues, open a thread or contribute directly to the Ultralytics GitHub.

Big thanks to every community member and contributor—these advancements are only possible thanks to your ongoing support and collaboration!

Happy tracking!
— The Ultralytics Team