OxUvA
收藏资源简介:
OxUvA数据集是由牛津大学等机构创建,专注于评估单目标跟踪算法,特别强调长序列和目标频繁消失的情况。该数据集包含366个序列,总时长超过14小时,是当前最大的跟踪数据集。数据集内容涵盖多种视频场景,旨在解决传统短序列数据集无法代表实际应用需求的问题。创建过程中,通过精心筛选和标注,确保数据集的质量和多样性。OxUvA数据集的应用领域广泛,包括视频分析、监控、机器人技术、增强现实和视频编辑,旨在推动跟踪技术在复杂环境中的应用。
The OxUvA dataset was developed by the University of Oxford and other institutions, focusing on evaluating single-object tracking algorithms, with particular emphasis on long sequences and scenarios where targets frequently disappear. It contains 366 sequences with a total duration of over 14 hours, making it the largest tracking dataset currently available. The dataset covers diverse video scenarios, aiming to address the issue that traditional short-sequence datasets fail to represent real-world application requirements. During its creation, rigorous screening and annotation were carried out to ensure the quality and diversity of the dataset. The OxUvA dataset has a wide range of application fields, including video analysis, surveillance, robotics, augmented reality and video editing, and is intended to promote the application of tracking technology in complex environments.

- OxUvA数据集首次发布,由牛津大学提出,旨在评估视频分类任务中的长期行为识别能力。
- OxUvA数据集首次应用于国际计算机视觉与模式识别会议(CVPR)的挑战赛中,推动了视频分析领域的研究进展。
- OxUvA数据集的扩展版本发布,增加了更多的视频样本和类别,进一步丰富了数据集的内容和多样性。
- OxUvA数据集在多个国际顶级会议上被广泛引用和讨论,成为视频行为识别领域的重要基准数据集之一。
- OxUvA数据集的应用范围扩展至深度学习和人工智能的其他领域,如自动驾驶和智能监控系统。
- 1Long-term Recurrent Convolutional Networks for Visual Recognition and DescriptionUniversity of Oxford · 2015年
- 2Temporal Segment Networks: Towards Good Practices for Deep Action RecognitionUniversity of Oxford · 2016年
- 3A Closer Look at Spatiotemporal Convolutions for Action RecognitionFacebook AI Research · 2018年
- 4Temporal Relational Reasoning in VideosUniversity of Oxford · 2018年
- 5SlowFast Networks for Video RecognitionFacebook AI Research · 2019年



