Fine-Grained Vehicle Detection (FGVD) Dataset
收藏资源简介:
FGVD数据集是由印度理工学院德里分校和海得拉巴国际信息技术研究所创建的,旨在为不受限制的道路环境提供细粒度的车辆检测。该数据集包含5502张场景图像,拥有210个独特的细粒度标签,涵盖多种车辆类型,并采用三级层次结构组织。数据集的创建过程涉及从IDD检测数据集中选择高质量图像,并通过专业的标注团队进行精细标注。FGVD数据集的应用领域包括智能交通监控系统,用于车辆再识别和在密集及遮挡交通场景中的鲁棒检测,解决现有检测模型在处理复杂交通场景时的不足。
The FGVD dataset was developed by the Indian Institute of Technology Delhi and the International Institute of Information Technology Hyderabad, with the goal of facilitating fine-grained vehicle detection in unconstrained road environments. It comprises 5502 scene images, featuring 210 unique fine-grained labels covering a wide range of vehicle types, and is structured under a three-level hierarchical framework. The construction of this dataset entails selecting high-quality images from the IDD detection dataset and performing precise annotation via a professional annotation team. The FGVD dataset finds applications in intelligent traffic monitoring systems, supporting vehicle re-identification and robust detection in dense and occluded traffic scenarios, and addresses the limitations of existing detection models when dealing with complex traffic scenes.




