TUM monoVO dataset
收藏资源简介:
TUM monoVO数据集是由慕尼黑工业大学创建,用于评估单目视觉测距和SLAM方法的跟踪精度。该数据集包含50个真实世界的序列,总计超过100分钟的视频,涵盖从狭窄的室内走廊到宽阔的户外场景等多种环境。所有序列主要包含探索性相机运动,起点和终点相同,便于通过累积漂移评估跟踪精度。数据集特别之处在于所有序列都进行了光度校准,提供了每帧的曝光时间、相机响应函数和密集镜头衰减因子。此外,数据集还提出了一种新颖的非参数渐晕校准方法,简化了校准设置。该数据集主要应用于解决自主车辆、四旋翼飞行器到虚拟和增强现实等领域中的视觉测距和地图构建问题。
The TUM monoVO dataset was created by the Technical University of Munich (TUM) to evaluate the tracking accuracy of monocular visual odometry and SLAM methods. This dataset contains 50 real-world sequences, totaling over 100 minutes of video, covering diverse environments ranging from narrow indoor corridors to expansive outdoor scenes. All sequences primarily feature exploratory camera motions, with identical start and end points, which facilitates tracking accuracy assessment via accumulated drift. A notable feature of this dataset is that all sequences have undergone photometric calibration, providing exposure time, camera response function, and dense lens attenuation factors for each frame. Additionally, the dataset proposes a novel non-parametric vignetting calibration method that simplifies the calibration setup. This dataset is mainly applied to solve visual odometry and mapping problems in fields such as autonomous vehicles, quadrotors, virtual reality (VR) and augmented reality (AR).

- 1A Photometrically Calibrated Benchmark For Monocular Visual Odometry慕尼黑工业大学 · 2016年



