ultralytics 8.3.170

v8.3.170 - `ultralytics 8.3.170` Eliminate `batch` argument for `AutoBackend` class and fix potential batch inconsistency for OpenVINO (#21452)

on Python PyPI

🌟 Summary

This release streamlines batch processing and inference for Ultralytics models—especially with OpenVINO—while boosting reliability, transparency, and user experience across the platform. 🚀

📊 Key Changes

• Batch Handling Overhaul: Removed the need to manually set batch size for model backends; batch size and dynamic inference settings are now automatically read from model metadata.
• OpenVINO Improvements:
  • Smarter selection of inference modes for OpenVINO, optimizing performance for different batch and dynamic settings.
  • Enhanced logging now shows which device (CPU, GPU, etc.) is used for OpenVINO inference.
  • Improved support for batch inference with OpenVINO-exported models, ensuring correct output structure.
• Image Loading Robustness: The image reading utility now safely handles unreadable or corrupted images, preventing crashes.
• Documentation Upgrade: Embedded a YouTube video in the YOLO data augmentation guide, offering a visual walkthrough of techniques like Mosaic and MixUp.

🎯 Purpose & Impact

• 📦 Simplified Setup: Users no longer need to worry about batch size configuration—it's handled automatically, reducing setup errors and confusion.
• ⚡ Optimized Performance: Models, especially those exported to OpenVINO, now run more efficiently and adaptively, making the most of available hardware and settings.
• 🖥️ Better Transparency: Clearer logs help users understand which devices are running their models, aiding in debugging and performance tuning.
• 🛡️ Increased Reliability: Improved error handling ensures smoother workflows, even with problematic image files.
• 🎥 Enhanced Learning: The new video resource in the docs makes advanced data augmentation techniques more accessible to all users.

Overall, this update makes Ultralytics models easier to use, more robust, and better documented—helping both new and experienced users get the most out of their AI projects! 🌟

What's Changed

• Fixing NoneType error while reading image by @nikunjlad in #21467
• Added info about used device in OpenVINO by @adrianboguszewski in #21498
• Add https://youtu.be/e-TwqFtay90 to docs by @RizwanMunawar in #21511
• Fix OpenVINO segment inference when batch>1 by @Laughing-q in #21508
• ultralytics 8.3.170 Eliminate batch argument for AutoBackend class and fix OpenVINO inference by @Laughing-q in #21452

New Contributors

• @nikunjlad made their first contribution in #21467

Full Changelog: v8.3.169...v8.3.170