Penng13/EgoExOR

EgoExOR是一个结合第一人称和第三人称视角的手术室数据集，旨在全面理解手术活动。数据集包含94分钟的两类模拟脊柱手术（超声引导针插入和微创脊柱手术）的数据，整合了来自可穿戴眼镜的自我中心数据（RGB、视线、手部追踪、音频）、来自RGB-D相机的第三人称RGB和深度数据，以及超声图像。数据集提供了详细的场景图注释，包含36个实体和22种关系（568,235个三元组），支持临床交互的鲁棒建模，适用于动作识别和以人为中心的感知等任务。数据集还评估了两种最先进模型的手术场景图生成性能，并提供了一个新的基线，明确利用了EgoExOR的多模态和多视角信号。

EgoExOR is an egocentric–exocentric operating room dataset designed for comprehensive understanding of surgical activities. It spans 94 minutes of two emulated spine procedures, Ultrasound-Guided Needle Insertion and Minimally Invasive Spine Surgery, integrating egocentric data (RGB, gaze, hand tracking, audio) from wearable glasses, exocentric RGB and depth from RGB-D cameras, and ultrasound imagery. The dataset includes detailed scene graph annotations covering 36 entities and 22 relations (568,235 triplets), enabling robust modeling of clinical interactions for tasks like action recognition and human-centric perception. It also evaluates the surgical scene graph generation performance of two adapted state-of-the-art models and offers a new baseline that explicitly leverages EgoExOR’s multimodal and multi-perspective signals.