AcinoSet

AcinoSet是由非洲机器人单元创建的一个大型3D姿态估计数据集，专注于野生猎豹的复杂动态。该数据集包含119,490帧多视角同步高速视频，以及7,588帧人工标注的2D关键点。数据收集自南非的两个研究中心，使用六台GoPro相机进行拍摄。AcinoSet的创建旨在为生态学、神经科学、机器人学、生物力学及计算机视觉等领域提供一个强大的3D基准，以促进对猎豹等高度灵活动物的研究。AcinoSet is a large-scale 3D pose estimation dataset developed by the African Robotics Unit, focusing on the complex dynamics of wild cheetahs. This dataset includes 119,490 frames of multi-view synchronized high-speed video, as well as 7,588 frames of manually annotated 2D keypoints. The data was collected from two research centers in South Africa, captured using six GoPro cameras. AcinoSet was created to serve as a robust 3D benchmark across disciplines including ecology, neuroscience, robotics, biomechanics, and computer vision, facilitating research on highly agile animals such as cheetahs.

搜集汇总

背景与挑战

背景概述

AcinoSet是一个专注于野生猎豹的3D姿态估计数据集，包含119,490帧多视角同步高速视频、相机校准文件和7,588帧人工标注的2D关键点。该数据集提供了三种基线方法（稀疏束调整、扩展卡尔曼滤波和全轨迹估计）用于3D姿态估计，并附带3D轨迹和交互工具，适用于生态学、机器人学、生物力学和计算机视觉等多个领域的研究。