Luka-Wang/compositional_control

该数据集是一个多模态数据集，包含图像、姿态条件图像、分割条件图像和文本描述。具体特征包括：image（原始图像）、pose_condition（作为条件的姿态图像）、seg_condition（作为条件的分割图像）和caption（与图像相关的文本描述）。数据集仅包含验证集（val），共有1154个样本，总大小约为335MB，下载大小约为334MB。数据文件路径为data/val-*。该数据集可能用于计算机视觉任务，如图像生成、姿态估计或分割，但README未提供具体描述，因此基于特征推断其潜在应用。

This dataset is a multimodal dataset containing images, pose condition images, segmentation condition images, and text captions. The features include: image (original image), pose_condition (pose image used as a condition), seg_condition (segmentation image used as a condition), and caption (text description associated with the image). The dataset only includes a validation split (val) with 1,154 examples, total size approximately 335MB, and download size approximately 334MB. Data files are located at data/val-*. The dataset is likely intended for computer vision tasks such as image generation, pose estimation, or segmentation, but no specific description is provided in the README, so this is inferred from the features.