Convert the COCO RLE format to YOLOv5/v8 segmentation format.

[ryouchinsa]

This PR fixes these problems.

Convert the COCO RLE format to YOLOv5/v8 segmentation format.
#38

Converting the COCO keypoints format to YOLOv8 pose format.
#45

copilot:all

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

📊 Key Changes

• Added support for COCO RLE (Run Length Encoding) to YOLOv5/v8 segmentation format conversion.
• New use_keypoints parameter introduced to handle keypoints annotation conversion.
• Enhanced support for multi-segment continuous instance merging and proper shape detection.
• Various utility functions added for handling keypoints and segmentations, such as bbox_from_keypoints, show_kpt_shape_flip_idx, merge_contours, and rle2polygon.

🎯 Purpose & Impact

The PR allows users to convert COCO dataset annotations that include RLE formatted segmentations and keypoints to a YOLO compatible format. This opens up the capability for users to train YOLO models with datasets that have richer annotations, potentially leading to more precise object detection, especially for tasks that require exact outline (segmentation) or specific point detection (keypoints).

🌟 Summary

Now you can convert COCO's detailed RLE and keypoints annotations to YOLO format seamlessly! 🎨🔑