Muddy Racer re-IDentification Dataset (MUDD)

MUDD数据集是首个针对越野摩托车赛手在极端条件下重新识别的大型基准数据集。该数据集包含3906张图像，涵盖150个身份，由16名专业摩托车运动摄影师在10场越野赛事中拍摄。数据集特点包括严重的泥浆遮挡、运动模糊、复杂姿态和极端光照条件，这些都是在现有重新识别数据集中未见过的。创建过程中，采用了结合辅助信息的注释方法，显著提高了标注效率。MUDD数据集主要用于推动在不受控制的现实世界条件下进行重新识别的研究，特别是在新兴的运动分析领域。

The MUDD dataset is the first large-scale benchmark dataset for re-identification of off-road motorcycle racers under extreme conditions. Comprising 3906 images spanning 150 identities, it was captured by 16 professional motorsports photographers across 10 off-road racing events. Notable characteristics of the dataset include heavy mud occlusion, motion blur, complex poses, and extreme lighting conditions, all of which are unprecedented in existing re-identification datasets. During its creation, an annotation method incorporating auxiliary information was adopted, which significantly improved annotation efficiency. The MUDD dataset is primarily intended to advance re-identification research under uncontrolled real-world conditions, particularly in the emerging field of sports analytics.