ljnlonoljpiljm/blip3-grounding-50m

这是一个包含图像信息的数据集，每个图像都由一个唯一的uid标识，并附带url和图像描述caption。每个图像还包含了检测到的对象的数量num_detections和具体的检测结果detections。检测结果包括对象的标签label、边界框box和置信度confidence。数据集分为训练集train，共有约5237万个示例。

This is a dataset containing image information, each image is identified by a unique uid, and is accompanied by a url and an image description caption. Each image also contains the number of detected objects num_detections and specific detection results detections. The detection results include the objects label, bounding box box, and confidence score. The dataset is split into a training set train, with a total of about 52.37 million examples.