YesBut
收藏资源简介:
YesBut数据集由印度理工学院卡拉格普尔分校创建,旨在评估视觉语言模型对讽刺理解的能力。该数据集包含2547张图像,其中1084张为讽刺图像,1463张为非讽刺图像,涵盖多种艺术风格。数据集的创建过程包括从社交媒体收集图像、人工标注、使用DALL-E 3生成2D和3D图像等步骤。YesBut数据集主要应用于多模态任务,如讽刺图像检测、理解和完成,旨在解决现有视觉语言模型在讽刺理解上的不足。
The YesBut Dataset was developed by the Indian Institute of Technology Kharagpur to evaluate the ability of vision-language models to understand sarcasm. This dataset contains 2,547 images in total, including 1,084 sarcastic images and 1,463 non-sarcastic images, covering a diverse range of art styles. The construction process of the dataset includes collecting images from social media, manual annotation, and generating 2D and 3D images using DALL-E 3. The YesBut Dataset is mainly applied to multimodal tasks such as sarcastic image detection, understanding and completion, aiming to address the limitations of existing vision-language models in sarcasm comprehension.




