New Release: Ultralytics v8.2.85

[glenn-jocher]

Ultralytics YOLO v8.2.85 Release Announcement

Summary

We are excited to announce the release of Ultralytics YOLO v8.2.85! This update brings significant enhancements, including new features, improved workflows, and better compatibility across the platform. Dive into the details below to see what’s new and how it can benefit your projects.

Key Changes

YOLOv10 Parameter Fix

• New max_det Parameter: This update introduces the max_det parameter, allowing users to specify the maximum number of detections. This enhancement is a crucial step towards the upcoming YOLOv10, providing greater control and customization over model outputs.
  • PR by @ambitious-octopus

GitHub Actions Update

• Streamlined Publish Workflow: We have simplified the publish workflow by removing the openai dependency and consolidating complex scripts into a single command. This change reduces complexity and potential dependency issues, facilitating faster and more efficient publishing of updates.
  • PR by @glenn-jocher
  • PR by @glenn-jocher

INT8 Export Warning

• Automatic dynamic=True Enforcement: Added a warning for automatic enforcement of dynamic=True during INT8 model exports to enhance compatibility, ensuring smoother user experiences with advanced export settings.
  • PR by @Burhan-Q

Documentation Enhancements

• Improved Author Attribution: Enhanced documentation with author avatars, optimizing related tools and making contributions more visible and accessible to users.
  • PR by @glenn-jocher

Explorer Tests Requirement

• PyTorch Version Update: Updated tests to require PyTorch version 1.13 or newer, ensuring compatibility, reliability, and stability across development environments.
  • PR by @glenn-jocher

Purpose & Impact

• YOLOv10 Readiness: The new max_det parameter is a key enhancement that sets the stage for YOLOv10, giving users greater control and customization options over model outputs.
• Optimized Release Workflow: The streamlined workflow reduces complexity and potential dependency issues, facilitating faster and more efficient publishing of updates.
• Enhanced Export Compatibility: The INT8 export warning for dynamic=True ensures compliance with optimal settings, improving export reliability and performance.
• Improved Documentation: Enhanced visualization with author avatars increases transparency and improves user interaction with the documentation.
• Reliable Testing: By enforcing minimum version requirements for PyTorch, the update guarantees stable and consistent testing, enhancing overall development reliability.

These updates emphasize Ultralytics’ commitment to enhancing YOLOv10’s functionality, improving user control, and refining the overall development and deployment experience.

What’s Changed

• Optimize docs author avatars by @glenn-jocher in PR #15923
• Explorer tests require torch>=1.13 by @glenn-jocher in PR #15930
• Update notebooks: Fix classes_names argument with names by @RizwanMunawar in PR #15915
• Includes warning for enforced dynamic during INT8 exports by @Burhan-Q in PR #15874
• Update publish.yml by @glenn-jocher in PR #15932
• Update publish.yml by @glenn-jocher in PR #15933
• ultralytics 8.2.85 YOLOv10 max_det arg fix by @ambitious-octopus in PR #15917

Full Changelog: v8.2.84…v8.2.85

Try It Out!

We encourage you to try out the new release and provide feedback. Your input is invaluable in helping us improve and deliver the best possible experience. Download the latest version here.

Thank you for your continued support and contributions to the Ultralytics community!