Open-Ended PitVQA

该数据集包含了大约101,803帧来自25个程序性手术视频，以及745,972个涵盖关键手术要素的问答句子对。此外，该数据集还涉及垂体手术的各个方面，包括工具的检测与定位。规模上，数据集包含了101,803帧画面和745,972个问答对，其任务主要集中在视觉问答领域。

This dataset contains approximately 101,803 frames from 25 procedural surgical videos and 745,972 question-answer sentence pairs covering key surgical elements. It covers all aspects of pituitary surgery, including tool detection and localization. With a total scale of 101,803 frames and 745,972 question-answer pairs, the core tasks of this dataset primarily focus on the visual question answering (VQA) domain.

搜集汇总

背景与挑战

背景概述

Open-Ended PitVQA数据集是一个专注于医学内窥镜垂体手术的视觉问答数据集，包含25个手术视频的101,803个图像帧和745,972个问答对。该数据集基于专业手术标注框架，涵盖手术阶段、步骤和器械等开放性问题，适用于视觉问答模型训练和评估。