Monocular Inertial Visual Object Tracking (MIVOT)

The Monocular Inertial Visual Object Tracking (MIVOT) dataset is a novel dataset designed for 3D object detection and tracking from monocular and inertial inputs, specifically for dynamic environments like UAV applications. It includes synchronized RGB, inertial data, and ground truth depth annotations recorded with a RealSense D455i and Ouster OS0 LiDAR, simulating UAV perspectives with challenging scenarios such as bird's-eye views and dynamic movements, making it suitable for evaluating advanced object detection and tracking frameworks.单目惯性视觉目标跟踪（Monocular Inertial Visual Object Tracking，MIVOT）数据集是一款专为基于单目与惯性输入的三维目标检测与跟踪任务打造的新型数据集，尤其适配无人机（Unmanned Aerial Vehicle，UAV）应用这类动态场景。该数据集包含由RealSense D455i与Ouster OS0激光雷达（LiDAR）采集的同步RGB图像、惯性数据与真值深度标注，通过模拟无人机视角构建了涵盖鸟瞰视角与动态运动等在内的极具挑战性的场景，可用于评估先进的目标检测与跟踪算法框架。