BEOID (Bristol Egocentric Object Interactions Dataset)

BEOID 数据集包括从准备咖啡到操作举重机和开门的对象交互。数据集记录在六个位置：厨房、工作区、激光打印机、带锁门的走廊、心脏健身房和举重机。对于前四个位置，记录了来自五个不同操作员的序列（每个操作员两个序列），并记录了最后两个位置的三个操作员（每个操作员三个序列）。可穿戴凝视跟踪器硬件（ASL Mobile Eye XG）用于记录数据集。提供具有校准的 2D 凝视注视的同步宽镜头视频数据。此外，我们使用预先构建的云点图和 PTAM 跟踪来发布 3D 信息。包括图像和注视的三维信息。

The BEOID dataset consists of object interactions ranging from preparing coffee and operating weightlifting machines to opening doors. The dataset was recorded at six locations: kitchen, workspace, laser printer, corridor with locked doors, cardiac gym, and weightlifting machine. For the first four locations, sequences from five distinct operators were recorded, with two sequences per operator; for the final two locations, sequences from three operators were recorded, with three sequences per operator. The dataset was captured using wearable gaze tracker hardware (ASL Mobile Eye XG). Synchronized wide-angle video data with calibrated 2D gaze fixation is provided. Additionally, 3D information is released using pre-built point cloud maps and PTAM tracking. This dataset includes 3D information incorporating both images and gaze fixation data.

搜集汇总

背景与挑战

背景概述

BEOID是一个第一视角对象交互数据集，记录6种日常场景下的物体操作行为，包含同步的2D凝视数据和3D空间信息。数据集由布里斯托尔大学于2014年发布，使用可穿戴式眼动仪采集，涵盖厨房、健身房等多种环境。