noahschaffer/clasp-audioset

该数据集包含图像及其对应的文本描述，每个样本还包括标签列表、YouTube视频ID和起始时间戳。数据集分为训练集（17,372个样本）和评估集（15,778个样本），图像和描述用于多模态学习任务。This dataset contains images paired with textual captions, along with a list of labels, YouTube video ID, and start timestamp. It is split into training (17,372 examples) and evaluation (15,778 examples) sets, intended for multimodal learning tasks.