jomoll/silent-heart-dataset-p005

该数据集包含了图片及其相关描述，每个样本都包括一个唯一的标识符、图片、文本描述、描述的ID、图片的高度分类标签（decoy或heart）、背景编号、心率、左心室率、诱饵形状、命名物体、静音物体、特殊物体、是否提及心脏的布尔值和目标概率。数据集分为训练集、验证集和测试集，分别包含400000、4000和4000个样本。

The dataset consists of images and their corresponding captions. Each sample includes a unique identifier, an image, a textual description, the ID of the caption, a label for image height classification (decoy or heart), background number, heart rate, left ventricular rate, decoy shape, named objects, silent objects, special object, a boolean indicating if the heart is mentioned, and the probability of the target. The dataset is split into training, validation, and test sets, containing 400,000, 4,000, and 4,000 samples respectively.