Fix min_items

[glenn-jocher]

Signed-off-by: Glenn Jocher glenn.jocher@ultralytics.com

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Optimization of image filtering logic in data loader.

📊 Key Changes

• Updated the conditional check from len(x) > min_items to len(x) >= min_items for determining which images to include in a dataset.
• Fixed the logging message to correctly display the number of images filtered from the dataset using n instead of nf.

🎯 Purpose & Impact

• The modification ensures images with exactly min_items number of annotations will also be included in the dataset, rather than strictly more than min_items.
• The updated logging message provides accurate feedback on the dataset filtering process, aiding in debugging and dataset understanding.
• 🧑‍💻 Developers will benefit from more precise dataset creation, and 📈 model training may improve due to the inclusion of additional valid images.
• 🖼️ Users leveraging the Ultralytics YOLOv5 framework for object detection tasks may experience slightly increased dataset sizes and potentially more robust training results.