SAR-RARP50

该数据集是一个针对开放式前列腺根治术中缝合任务期间错误检测的体内手术错误标注数据集。它采用了临床上验证过的观察性临床人员可靠性评估工具（OCHRA）进行帧级别的标注。数据集的规模包括40个训练视频，其中包含46,586个正常帧和18,540个错误帧；以及8个测试视频，包含7,549个正常帧和2,096个错误帧。该数据集的任务是进行外科手术错误检测。

This dataset is an in-vivo surgical error annotated dataset for error detection during the suturing task of open radical prostatectomy. It was annotated at the frame level using the clinically validated Observational Clinical Human Reliability Assessment (OCHRA) tool. The dataset includes 40 training videos containing 46,586 normal frames and 18,540 error frames, as well as 8 test videos with 7,549 normal frames and 2,096 error frames. The task of this dataset is surgical error detection.

搜集汇总

背景与挑战

背景概述

SAR-RARP50数据集是一个专门用于机器人辅助手术错误检测的数据集，包含帧级错误标注和DINOv2提取的空间嵌入序列。该数据集支持高效的长序列建模，旨在提高手术错误检测的准确性和计算效率。