Racer Number Dataset (RND) 和 Muddy Racer re-iDentification Dataset (MUDD)

本研究介绍了两个新的挑战性真实世界数据集：Racer Number Dataset (RND) 和 Muddy Racer re-iDentification Dataset (MUDD)，旨在突出当前方法的不足并推动极端条件下OCR和人员再识别（ReID）的进步。这两个数据集包含超过6,300张在越野比赛中拍摄的图像，展示了多种因素，如泥浆、复杂姿态和运动模糊，这些因素甚至削弱了现代视觉系统的能力。RND包含2,411张图像，展示了参与比赛的骑手，每张图像都标注了可见骑手号码的边界框和数字文本转录。MUDD包含3,906张图像，捕捉了150名骑手在10个不同的越野赛事中的表现，每个骑手裁剪都标注了唯一的身份标签。这些数据集的应用领域包括精确的照片搜索、体育分析、自动计时和评分等，旨在解决在极端环境中准确识别骑手和检测赛车号码的问题。

This study introduces two novel challenging real-world datasets: the Racer Number Dataset (RND) and the Muddy Racer re-identification Dataset (MUDD), which aim to highlight the limitations of current approaches and advance the development of OCR and person re-identification (ReID) under extreme conditions. These two datasets contain over 6,300 images captured during off-road races, which feature various challenging factors such as mud, complex poses and motion blur—factors that can even degrade the performance of modern vision systems. RND consists of 2,411 images showing race riders, with each image annotated with bounding boxes and digital text transcriptions of the visible rider numbers. MUDD includes 3,906 images capturing 150 riders across 10 distinct off-road racing events, and each cropped rider image is annotated with a unique identity label. The application scenarios of these datasets cover precise photo search, sports analytics, automatic timing and scoring, among others, and they are designed to address the challenges of accurately identifying riders and detecting race numbers in extreme environments.