V3Det

V3Det（超大类别目标检测数据集）精确标注了超过13,029类物体，助力更通用的目标检测研究。 1) 超大类别：V3Det涵盖了在真实图像上超过13,000类物体的检测框标注。 2）层级的类别结构：V3Det标注了类别的从属关系，形成了一颗完善的类别关系树。 3）丰富标注：V3Det拥有24.5万张标注图片和由人类专家和chatgpt提供的超过13,000个类别的专业介绍。V3Det (Ultra-large-category Object Detection Dataset) accurately annotates over 13,029 object categories, facilitating more generalizable object detection research. 1) Ultra-large category coverage: V3Det provides bounding box annotations for over 13,000 object categories in real-world images. 2) Hierarchical category structure: V3Det annotates the subordinate relationships between categories, forming a complete category relationship tree. 3) Rich annotations: V3Det contains 245,000 annotated images and professional introductions for over 13,000 categories provided by human experts and ChatGPT.