COMPASS

COMPASS数据集是由弗吉尼亚大学开发的一个综合性的外科手术任务模型和数据集。该数据集包含了来自三个公开数据集（JIGSAWS, DESK, ROSMA）的六个干实验室外科任务，涵盖了运动学和视频数据以及上下文和运动原语标签。数据集的创建旨在通过统一的动素集来实现更客观的标签化和不同数据集的聚合，从而支持外科手术过程分析、技能评估、错误检测和自主性等应用。

The COMPASS dataset is a comprehensive surgical task model and dataset developed by the University of Virginia. It includes six dry-laboratory surgical tasks sourced from three publicly available datasets (JIGSAWS, DESK, ROSMA), covering kinematic and video data alongside contextual and motion primitive labels. The dataset was created to enable more objective labeling and cross-dataset aggregation via a unified set of motion primitives, thereby supporting applications such as surgical workflow analysis, skill assessment, error detection, and autonomy.