New Release: Ultralytics v8.3.54

Discussion
,
1

:rocket: New Release: Ultralytics v8.3.54

We’re thrilled to announce the release of Ultralytics v8.3.54, packed with powerful new features and enhancements to streamline your workflows and elevate real-time inference and model deployment experiences!

:star2: Summary

This update introduces a revamped Streamlit-based real-time inference solution, better flexibility for OpenVINO model exports, and improved YOLOv11 documentation. Additionally, it includes dependency updates for smoother development and showcases a new ONNXRuntime deployment example for RTDETR. Let’s dive into the details!

:bar_chart: Key Changes

:rocket: Revamped Streamlit Inference Tool

  • New Inference class designed for smoother real-time operations.
  • Features a sidebar for:
    • Quick configuration of video sources, models, and settings like confidence thresholds.
    • Seamless webcam or video upload support for live predictions.
  • Enhanced interactivity:
    • Class selection, FPS monitoring, and object tracking integrated into the tool.

:package: OpenVINO Export Enhancements

  • Introduced dynamic shape support, boosting deployment versatility for diverse scenarios.
  • Standardized export arguments for formats including batch and dynamic.

:open_book: YOLOv11 Documentation Updates

  • Revised guides to reflect YOLOv11 region-based object counting, ensuring users have the most up-to-date resources.

:globe_with_meridians: ONNXRuntime Example for RTDETR

  • Added a practical example to deploy RTDETR models with ONNXRuntime in Python, reducing entry barriers for developers.

:snake: Python Workflow Updates

  • Minimum Python version for CI workflows updated to 3.9, aligning with the latest ecosystem standards.

:gear: Dependency Updates

  • Upgraded GitHub Actions setup-uv workflow to v5, improving caching and build efficiency.

:dart: Why This Matters

  • Ease of Use: Real-time inference with Streamlit is now simpler and more intuitive for users of all backgrounds.
  • Deployment Flexibility: Dynamic OpenVINO exports ensure seamless scaling across platforms.
  • Clarity for YOLOv11 Users: Accurate documentation supports precise implementation in custom workflows.
  • Dev-Friendly Updates: From ONNXRuntime examples to upgraded CI workflows, development processes are more robust than ever!

:link: What’s Changed

Here’s the list of key updates, along with the contributors who made it all possible:

  1. Add dynamic to approved OpenVINO export args – by @glenn-jocher
    #18353
  2. Bump astral-sh/setup-uv from 4 to 5 in /.github/workflows – by @dependabot[bot]
    #18358
  3. Update YOLOv8 to YOLO11 in region-counting.md – by @RizwanMunawar
    #18360
  4. Min CI Python 3.9 from 3.8 – by @glenn-jocher
    #18355
  5. [Example] RTDETR-ONNXRuntime-Python – by @semihhdemirel
    #18369
  6. ultralytics 8.3.54 New Streamlit inference Solution – by @RizwanMunawar
    #18316

Full Changelog: v8.3.53…v8.3.54

:raised_hands: Get Involved

We encourage you to try out the new release today and let us know your thoughts! Your feedback is invaluable and helps drive future improvements.

Check out the full release here: v8.3.54

Thank you for being an amazing part of the YOLO and Ultralytics community! Your continuous support and engagement inspire us to innovate and improve every step of the way. Happy inferencing! :mag: