EgoSurgery-HTS
收藏资源简介:
EgoSurgery-HTS数据集是由庆应义塾大学创建的,包含21段时长总计15小时的手术视频,由8位医生完成。该数据集在EgoSurgery的基础上增加了手术工具、手部以及手工具交互的像素级注释,共包含19496张高质量图像,并对50383个工具、57173只手以及41605个手工具交互进行了标注。数据集旨在推动开放手术视频中egocentric场景的精细理解,定义了三种任务:手部分割、工具分割和手工具分割,以促进对egocentric开放手术场景的理解。
The EgoSurgery-HTS dataset was created by Keio University. It consists of 21 surgical videos with a total duration of 15 hours, conducted by 8 surgeons. Building upon the original EgoSurgery dataset, this dataset introduces pixel-level annotations for surgical instruments, hands, and hand-tool interactions, encompassing a total of 19,496 high-quality images. Specifically, there are 50,383 instrument annotations, 57,173 hand annotations, and 41,605 hand-tool interaction annotations. The dataset aims to advance fine-grained understanding of egocentric scenes in open surgical videos, and defines three tasks to facilitate the understanding of egocentric open surgical scenarios: hand segmentation, instrument segmentation, and hand-tool segmentation.

- 1EgoSurgery-HTS: A Dataset for Egocentric Hand-Tool Segmentation in Open Surgery Videos庆应义塾大学 · 2025年



