NWPU-MOC
收藏资源简介:
NWPU-MOC数据集是由西北工业大学创建的一个大规模多类别目标计数数据集,包含3416个场景,每个场景分辨率为1024×1024像素,并精细标注了14个细粒度目标类别。该数据集不仅包含RGB图像,还包含近红外(NIR)图像,后者能提供更丰富的特征信息,有助于解决RGB图像中由于植被、光照和天气条件导致的可见性问题。NWPU-MOC数据集主要用于解决航空场景中的多类别目标计数问题,旨在通过密度图方法实现对不同类别目标的准确计数,并解决类别间的相互干扰问题。
The NWPU-MOC dataset is a large-scale multi-class object counting dataset developed by Northwestern Polytechnical University. It comprises 3416 scenes, each with a resolution of 1024 × 1024 pixels, and is meticulously annotated with 14 fine-grained object categories. This dataset includes not only RGB images but also near-infrared (NIR) images, which provide richer feature information and help resolve the visibility issues in RGB images caused by vegetation, illumination and weather conditions. The NWPU-MOC dataset is primarily designed for addressing multi-class object counting tasks in aerial scenarios, aiming to achieve accurate counting of objects across different categories via density map methods and mitigate mutual interference between different categories.




