ryushinn/11k-Hands-BBox-Keypoint

11k-Hands BBox Keypoint数据集包含为`ryushinn/11k-Hands`数据集估计的手部边界框和21点手部关键点。它是一个仅包含注释的辅助数据集，不复制源图像。数据集结构与源数据集相同，每行包含四个列：边界框坐标、边界框置信度、关键点坐标和关键点置信度。关键点的顺序遵循MMPose的COCO-WholeBody手部约定。注释是使用OpenMMLab模型生成的，包括RTMDet-Nano手部检测器和RTMPose-M Hand5手部姿态估计器。这些注释是模型估计的伪标签，而非人工标注的真实标签，适用于需要可重复手部边界框和手部姿态估计的研究工作流程。

The 11k-Hands BBox Keypoint dataset contains estimated hand bounding boxes and 21-point hand keypoints for the `ryushinn/11k-Hands` dataset. It is an annotations-only sidecar dataset that does not duplicate the source images. The dataset structure matches the source dataset, with each row containing four columns: bounding box coordinates, bounding box confidence score, keypoint coordinates, and keypoint confidence scores. The keypoints follow the COCO-WholeBody hand convention used by MMPose. Annotations were generated using OpenMMLab models, including the RTMDet-Nano hand detector and RTMPose-M Hand5 hand pose estimator. These annotations are model-estimated pseudo-labels, not manual ground-truth annotations, and are intended for research workflows requiring reproducible hand bounding boxes and pose estimates.