BURST
收藏资源简介:
BURST数据集是由亚琛工业大学和卡内基梅隆大学合作创建的,包含2914个多样化的视频,每个视频至少有480像素的分辨率,长度约30秒,涵盖室内外场景、野生视频、电影场景和车载街景等。数据集通过专业重新标注,提供了高质量的对象掩码,支持六个与视频中对象跟踪和分割相关的任务。BURST数据集旨在促进不同研究子社区之间的知识共享,加速通用方法的发展,以解决多个任务。
The BURST dataset was collaboratively developed by RWTH Aachen University and Carnegie Mellon University. It contains 2,914 diverse videos, each with a minimum resolution of 480 pixels and a duration of roughly 30 seconds, covering indoor and outdoor scenes, wildlife videos, film sequences, and in-vehicle street views, among other scenarios. The dataset provides high-quality object masks obtained via professional re-annotation, supporting six tasks related to object tracking and segmentation in videos. The BURST dataset is designed to promote knowledge sharing across different research sub-communities and accelerate the advancement of general-purpose methods for addressing multiple tasks.




