New Release: Ultralytics v8.3.108

Discussion
,
1

:rocket: Ultralytics v8.3.108 Release: YOLO11 IMX500 Export, New Guides, and Interactive Tracking UI!

We’re excited to announce the release of Ultralytics v8.3.108, packed with powerful new features, enhanced compatibility, performance improvements, and expanded learning resources. Whether you’re building edge AI applications or exploring advanced YOLO capabilities, this update brings practical tools and refinements for every user.

:glowing_star: Summary

Ultralytics v8.3.108 delivers native export support for YOLO11 models to Sony IMX500, improves interactive computer vision workflows, simplifies coding patterns, and introduces enhanced guides and educational resources.

:loudspeaker: New Features & Highlights

YOLO11 Sony IMX500 Export Support

  • Export Logic & Docs Updated: YOLO11 models are now fully supported for export to Sony IMX500 platforms, unlocking new possibilities for edge AI deployments including Raspberry Pi AI Cameras.
  • Benchmarking Results: YOLO11n models on IMX500 hardware display improved accuracy with minimal inference time impact.

YOLO11 Sony IMX export PR by @ambitious-octopus

Interactive Object Tracking UI Example

  • New Example App: Experiment with a hands-on UI for real-time object tracking using YOLO models. Features include click-to-track and visual overlays, making it perfect for demos, development, and learning.

YOLO-Interactive-Tracking-UI by @alireza787b

Enhanced Documentation & Educational Content

  • YOLO Data Augmentation Guide: A thorough new guide with visuals and explanations, helping you get the most out of your training data.
  • Project Goals Video Tutorial: YouTube tutorial now embedded directly in the docs for easy access.

New Data Augmentation Guide by @picsalex
YouTube Documentation Update by @RizwanMunawar

:hammer_and_wrench: Improvements

  • Simplified Codebase: Deprecated the crop_fraction parameter for classification transforms (PR by @fcakyon).
  • NUMPY Compatibility: Updated to use version-aware alternatives, future-proofing against NumPy deprecations (PR by @glenn-jocher).
  • Optimized Pose Angles: Pose angle calculations moved to use Python’s math module for consistency and performance.
  • Better Example Readability: Improved header organization and redirects in example README files (PR by @glenn-jocher).
  • Documentation Fixes: Indentation, formatting, and code alignment across several documentation pages (PRs by @RizwanMunawar).

:lady_beetle: Bug Fixes

  • Comet Error Fix: Corrects a validation issue during COCO dataset runs (PR by @Y-T-G).
  • Category Frequency Index Update: Added .squeeze() in category_freq to improve reliability (PR by @Laughing-q).
  • Solutions Test Disabled: Test stability improvements (PR by @RizwanMunawar).

:tada: New Contributors

Special welcome to @alireza787b for their first contribution with the interactive tracking UI example!

:books: Full Changelog & How to Get Started

:raising_hands: Try It & Share Your Feedback!

We encourage everyone to upgrade to v8.3.108, explore the new features and improvements, and share your experience. Your input is invaluable in shaping the future of Ultralytics and YOLO models!

As always, huge thanks to the entire YOLO community and Ultralytics contributors for your ongoing support and innovation.

Happy building!