Voxel51/WebUOT-238-Test
收藏Hugging Face2025-02-04 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/Voxel51/WebUOT-238-Test
下载链接
链接失效反馈官方服务：
资源简介：
WebUOT-238-Test是一个FiftyOne格式的数据集，包含238个样本，是从WebUOT-1M数据集中随机选取的视频，共有192,000+帧。WebUOT-1M是一个大规模的 underwater object tracking (UOT) 数据集，包含1,500个水下视频的1,100,000帧，涵盖408个目标类别，分为12个 superclass，包括鱼类、贝类和无生物等。数据集提供了高质量的边界框注释、23个跟踪属性和语言描述，适用于多模态跟踪研究。
WebUOT-238-Test is a FiftyOne dataset containing 238 samples, which are randomly selected videos from the WebUOT-1M dataset, totaling over 192,000 frames. WebUOT-1M is a large-scale underwater object tracking (UOT) benchmark with 1,100,000 frames across 1,500 underwater videos, covering 408 target categories categorized into 12 superclasses such as fish, molluscs, and inanimate objects. The dataset provides high-quality bounding box annotations, 23 tracking attributes, and language descriptions for multimodal tracking research.
提供机构：
Voxel51
