2017 Robotic Instrument Segmentation Challenge

机器人器械的分割是机器人辅助微创手术的一个重要问题。它可用于简单的 2D 应用，例如叠加遮罩或 2D 跟踪，也可用于更复杂的 3D 任务，例如姿势估计。在这个挑战中，我们邀请申请人参与 3 个不同的任务：二元分割、多标签分割和仪器识别。二进制分割仅涉及将图像分成仪器和背景，而多标签分割还要求用户识别仪器主体的哪些部分对应于达芬奇机器人仪器的不同关节部分。最后的识别任务测试用户能否识别出哪个分段对应哪个达芬奇仪器类型。为了实现这一点，我们提供了 8x 225 帧的机器人手术视频，以 2 Hz 的频率拍摄，Intuitive Surgical 的一个训练有素的团队手动标记了不同的部件和类型。邀请用户在作为测试集的 8x 75 帧视频和 2x 300 帧视频上测试他们的算法。描述来自：机器人仪器分割子挑战Robotic instrument segmentation is a critical problem in robot-assisted minimally invasive surgery. It can be applied to simple 2D applications such as overlay masks or 2D tracking, as well as more complex 3D tasks such as pose estimation. In this challenge, participants are invited to undertake three distinct tasks: binary segmentation, multi-label segmentation, and instrument recognition. Binary segmentation only involves dividing an image into instrument and background, while multi-label segmentation further requires participants to identify which parts of the instrument body correspond to the different joint sections of a da Vinci robotic instrument. The final recognition task tests whether participants can identify which segmented region corresponds to which type of da Vinci surgical instrument. To enable this, we provide 8 sets of 225-frame robotic surgery videos captured at 2 Hz, with manual annotations of different components and types completed by a well-trained team from Intuitive Surgical. Participants are invited to test their algorithms on the test sets consisting of 8 sets of 75-frame videos and 2 sets of 300-frame videos. Description source: Robotic Instrument Segmentation Sub-Challenge

搜集汇总

背景与挑战

背景概述

该数据集专注于机器人辅助微创手术中的器械分割，提供8x225帧标记视频，支持二元分割、多标签分割和仪器识别三个任务。数据集由麻省理工学院等机构于2019年发布，旨在促进手术机器人视觉算法的发展。