🌟 Summary
The v8.2.96 release of Ultralytics enhances data export capabilities, refines parking management solutions, and improves documentation workflows, offering better usability and integration for users.
📊 Key Changes
- Data Export Methods: Introduced new methods
to_df(),to_csv(), andto_xml()to export results in DataFrame, CSV, and XML formats. - Parking Management Optimization: Simplified the code and refactored data handling for easier setup and enhanced performance.
- Documentation and Streaming Updates: Improved documentation process and clarified streaming capabilities, including renaming and restructured examples for single and multi-stream video processing.
- Precision and Validation Enhancements: Aligned model validation precision with Automatic Mixed Precision settings for consistency.
🎯 Purpose & Impact
- Enhanced Exportability: Users can now easily export detection results in widely-used formats, facilitating better data analysis and integration with other tools. 📊
- Improved Clarity and Efficiency: The refactoring of parking management code enhances usability and processing speed, making it more intuitive for users managing parking data. 🚗
- Streamlined Documentation Workflow: Improved the process of auto-fixing and updating documentation, leading to more accurate and accessible developer resources. 📚
- Consistent Precision Handling: Ensures that validation and training are better aligned, leading to more reliable performance assessments and optimized use of computational resources. ⚙️
What's Changed
- Disable FP16 val on AMP fail and improve AMP checks by @Y-T-G in #16306
- Optimize
parking managementsolution by @RizwanMunawar in #16288 - Enable Docs auto-fixes on repo branches by @glenn-jocher in #16326
- Update Multi-Stream predict docs by @glenn-jocher in #16334
- Use
trainer.ampto determine FP16 validation by @Laughing-q in #16333 ultralytics 8.2.96newresults[0].to_dfPandas, XML and CSV methods by @MatthewNoyce in #16267
Full Changelog: v8.2.95...v8.2.96