SFU-HW-Tracks-v1

SFU-HW-Tracks-v1是由西蒙弗雷泽大学工程科学学院创建的数据集，专注于未压缩视频序列中的对象跟踪。该数据集包含13个HEVC CTC视频序列，每个序列均以YUV420格式提供，未经过压缩处理。数据集为每个视频帧提供了对象类别ID、对象ID以及边界框位置和尺寸的标注。创建过程中，数据集扩展了先前的SFU-HW-Objects-v1，为每个对象提供了唯一的标识符，使得能够追踪特定对象在帧间的移动。该数据集主要应用于研究视频压缩与对象跟踪之间的关系，以及评估对象跟踪算法的性能。

SFU-HW-Tracks-v1 is a dataset developed by the School of Engineering Science at Simon Fraser University, dedicated to object tracking in uncompressed video sequences. This dataset comprises 13 HEVC CTC video sequences, all supplied in YUV420 format with no compression applied. For every video frame, the dataset provides annotations including object category IDs, object IDs, as well as the positions and dimensions of bounding boxes. Built upon the earlier SFU-HW-Objects-v1, this dataset assigns unique identifiers to each individual object, allowing for the tracking of specific objects' trajectories across consecutive frames. This dataset is primarily utilized for researching the correlation between video compression and object tracking, as well as for evaluating the performance of object tracking algorithms.

搜集汇总

背景与挑战

背景概述

SFU-HW-Tracks-v1是由西蒙弗雷泽大学创建的专注于未压缩视频序列中对象跟踪的数据集，包含13个HEVC CTC视频序列，以YUV420格式提供，并提供了对象ID和边界框等标注，支持追踪对象在帧间的移动。该数据集主要用于研究视频压缩与对象跟踪的关系，以及评估对象跟踪算法的性能。