Transparent Object Tracking Benchmark (TOTB)
收藏资源简介:
Transparent Object Tracking Benchmark (TOTB) 是首个专门针对透明物体跟踪的基准数据集,由石溪大学计算机科学系创建。该数据集包含225个视频序列,总计86,000帧,涵盖15种不同类型的透明物体。每个视频序列均由人工标注,使用轴对齐的边界框。TOTB的创建旨在解决透明物体跟踪领域的挑战,特别是在机器人视觉和人类机器交互中的应用。数据集通过从YouTube收集原始视频并进行精细筛选和标注,确保了数据的质量和适用性。TOTB的应用领域广泛,旨在推动透明物体跟踪技术的发展,解决实际应用中的复杂问题。
Transparent Object Tracking Benchmark (TOTB) is the first dedicated benchmark dataset for transparent object tracking, developed by the Department of Computer Science at Stony Brook University. Comprising 225 video sequences with a total of 86,000 frames, the dataset covers 15 distinct types of transparent objects. Each video sequence is manually annotated using axis-aligned bounding boxes. The creation of TOTB aims to address the core challenges in the field of transparent object tracking, particularly for applications in robotic vision and human-machine interaction. The dataset is constructed by collecting raw videos from YouTube, followed by rigorous screening and annotation to ensure data quality and applicability. With a wide range of application scenarios, TOTB is designed to promote the advancement of transparent object tracking technologies and solve complex problems in real-world practical applications.




