AMP

AMP数据集是由卢布尔雅那大学计算机与信息科学学院创建的，专注于短期单目标视觉追踪的评估。该数据集包含17537帧的全景视频，每帧都进行了目标标注，并提供了十二种典型运动模式的生成器。创建过程中，视频主要从YouTube收集，并使用Ricoh Theta 360度相机录制额外序列。AMP数据集的应用领域主要集中在视觉追踪技术的性能评估，特别是在各种运动模式下的追踪表现，旨在解决现有基准测试中运动模式参数化不足的问题。

The AMP dataset was developed by the Faculty of Computer and Information Science, University of Ljubljana, with the core focus on evaluating short-term single-object visual tracking. This dataset consists of 17,537 frames of panoramic video footage, where each frame is equipped with target object annotations, and also offers generators for twelve distinct typical motion patterns. During its construction, the majority of the video sources were retrieved from YouTube, and supplementary sequences were recorded using a Ricoh Theta 360-degree camera. The primary application scenario of the AMP dataset lies in performance assessment of visual tracking techniques, specifically the tracking performance under varied motion patterns, which aims to tackle the shortcoming of inadequate parameterization of motion patterns in current benchmark test suites.