FELT
收藏资源简介:
FELT数据集是由安徽大学计算机科学与技术学院创建的,是目前最大的帧事件跟踪数据集,包含742个视频和1,594,474对RGB帧和事件流。该数据集旨在支持长期帧事件单目标跟踪任务,涵盖了45个目标对象类别和14个挑战属性。数据集的创建考虑了多模态跟踪的特殊性,特别是事件数据的空间稀疏性和RGB数据在弱化场景中的不完整性。FELT数据集的应用领域包括自动驾驶、无人机航拍和智能安全监控等,旨在解决长期跟踪中的频繁目标变化、噪声和不确定性问题。
The FELT dataset was developed by the School of Computer Science and Technology, Anhui University. It is currently the largest frame-event tracking dataset to date, containing 742 videos and 1,594,474 pairs of RGB frames and event streams. This dataset is designed to support long-term single-object tracking tasks that leverage both RGB frame and event modality data. It covers 45 target object categories and 14 challenging tracking attributes. The development of the FELT dataset takes into account the unique characteristics of multimodal tracking, particularly the spatial sparsity of event-based data and the incompleteness of RGB data in weakly illuminated scenarios. Application domains of the FELT dataset include autonomous driving, unmanned aerial vehicle (UAV) aerial photography, intelligent security monitoring and other related fields, aiming to address core challenges in long-term tracking including frequent target variations, environmental noise and inherent uncertainty.




