Ultralytics Release v8.3.35 Announcement
Summary
We’re excited to announce the release of Ultralytics v8.3.35! This update introduces significant advancements in model adaptability through enhanced support for dynamic input shapes in image processing. We’ve also revamped our Docker configuration for improved consistency and extended compatibility, while maintaining a focus on documentation and user experience improvements. Discover all the latest features and enhancements below!
Key Changes
New Features
- Dynamic Models Support: Our
pre_transformfunction now facilitates automatic letterboxing, streamlining your workflow when working with dynamic input shapes.
Improvements
- Updated Docker Configuration: Switched to Python 3.11.10 in Docker’s base image and included PaddlePaddle installation, broadening environment setup capabilities.
- Documentation Enhancements: Enhanced Ray Tune documentation, benchmarking tools, and introduced a scalable search bar for improved navigation on our documentation site.
Bug Fixes & Maintenance
- Cosmetic and Code Maintenance: Implemented JavaScript updates for a tidier codebase and refined styles to boost user interaction.
Purpose & Impact
- Enhanced Model Handling: The improvements in dynamic shape support empower users with more precise image preprocessing and potentially better model performance.
- Consistency and Compatibility: Docker’s enhanced configuration ensures smooth environment setups and supports a wider range of frameworks.
- Improved User Experience: Our comprehensive documentation updates, including a seamless search experience, help you find information more efficiently.
- Developer-Focused Refinements: Streamlined code and workflow updates support easier maintenance and development efficiency.
What’s Changed
- Addition of RTDETRv2 in
benchmarks.mdchart
by @RizwanMunawar - JavaScript updates in
extra.jsby @glenn-jocher - Docker base image updated to
ubuntu:latestby @glenn-jocher - Improvements to Docs Search Bar by @glenn-jocher
- Addition of missing import in Ray Tune docs by @Y-T-G
- ‘Auto’ letterbox support for dynamic models by @Laughing-q
Get Involved
We encourage everyone to try out the latest release and share your feedback! Your contributions help shape the future of our tools, and we value community input. Dive into the release details on GitHub and let us know your thoughts.
Thank you for being an amazing part of the Ultralytics community. Your support and feedback drive our progress!