Add min_items filter option

[glenn-jocher]

@AyushExel @Laughing-q dataset filter

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

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Implementation of a minimum item filter for dataset images in dataloaders.

📊 Key Changes

• Added a new parameter min_items to filter out images with fewer objects than specified.
• Improved error chaining in exceptions to help debugging.
• Implemented logic to filter images based on the number of annotated objects.

🎯 Purpose & Impact

• The min_items parameter allows users to tailor datasets to exclude images with too few objects, which can improve model training relevance and efficiency.
• Enhanced error messages aid users in troubleshooting dataset loading issues more easily.
• The change might impact dataset composition by potentially reducing the number of images available for training, but this should lead to more meaningful training if properly configured.