TImeD
收藏arXiv2025-09-30 收录
下载链接：
https://gaurav22verma.github.io/text-visualness/
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了3,620个英文句子，这些句子经过多位人类标注者的标注，附有视觉性评分。该数据集的主要用途是评估模型在区分视觉性文本与非视觉性文本方面的表现，以用于文本到图像检索和生成的视觉性检测任务。
This dataset consists of 3,620 English sentences, which have been annotated by multiple human annotators with corresponding visuality scores. Its primary purpose is to evaluate models' performance in distinguishing visual texts from non-visual texts, supporting visual detection tasks for text-to-image retrieval and generation.
