CityPersons
收藏资源简介:
CityPersons是一个基于Cityscapes数据集的新型行人检测数据集,由马克斯·普朗克信息学研究所创建。该数据集包含约35,000个高质量的行人边界框标注,覆盖了德国及邻近国家的27个城市。数据集的多样性体现在不同的季节、天气条件和人群密度,特别适合训练能够跨多个基准测试集泛化的单一CNN模型。CityPersons不仅用于训练更优秀的模型,还作为一个新的测试基准,尤其在处理重叠和多样环境方面更具挑战性。
CityPersons is a novel pedestrian detection dataset built upon the Cityscapes dataset, created by the Max Planck Institute for Informatics. It contains approximately 35,000 high-quality pedestrian bounding box annotations, covering 27 cities in Germany and its neighboring countries. The dataset features diversity across different seasons, weather conditions and crowd densities, making it particularly suitable for training single CNN models that can generalize across multiple benchmark datasets. Moreover, CityPersons can be used not only for training more advanced models, but also serves as a new testing benchmark, which is especially challenging when dealing with overlapping scenarios and diverse environments.




