Ego Motion Dataset

该数据集包含了超过1000段由互联行车记录仪获取的视频序列，用于训练视觉定位的自运动模型。每个视频大约40秒长，分辨率为1280x720。在训练自运动模型时，使用了这些视频中连续的两帧。该数据集的规模涵盖了大约1000辆不同车辆的1000段驾驶视频，其任务是进行自运动估计。

This dataset contains over 1000 video sequences acquired by connected driving dashcams, which are used to train egomotion models for visual localization. Each video is approximately 40 seconds long with a resolution of 1280×720. During the training of egomotion models, two consecutive frames from these videos are utilized. The dataset covers 1000 driving videos from approximately 1000 distinct vehicles, and its target task is egomotion estimation.