starvector/natural_wild_svg

该数据集包含了一系列字段，包括唯一的标识符(id)、SVG格式的图像描述(svg)、图像序列(images)、聚焦提示(focus_prompt)、描述(description)、真实图像(real_images)、边界框坐标序列(bbox)、一致性得分(constancy_score)和对齐得分(alignment_score)。数据集分为训练集(train)和评估集(eval)，其中训练集包含12759个示例，评估集包含1418个示例。数据集的总下载大小为4339708207字节，解压后大小为4648323453.63字节。根据这些信息，可以推测这是一个与图像处理相关的数据集，可能用于图像识别、图像描述生成等任务。

The dataset includes fields such as unique identifier (id), SVG image descriptions (svg), image sequences (images), focus prompts (focus_prompt), descriptions (description), real images (real_images), bounding box coordinate sequences (bbox), constancy scores (constancy_score), and alignment scores (alignment_score). It is split into a training set (train) with 12,759 examples and an evaluation set (eval) with 1,418 examples. The total download size of the dataset is 4,339,708,207 bytes, and the unpacked size is 4,648,323,453.63 bytes. Based on this information, it can be inferred that this is an image-related dataset, possibly used for tasks such as image recognition and image description generation.