VQG-23

VQG-23数据集是由印度理工学院卡拉格普尔分校开发的，用于少样本视觉问题生成任务。该数据集通过扩展原有的VQG-16数据集，增加了更多的图像-问题对和答案类别，总数达到299510条。创建过程中，数据集从Visual Genome数据集中提取并去除了重复图像，确保了数据的多样性和实用性。VQG-23数据集主要用于评估和推动少样本学习研究，特别是在视觉和语言生成任务中的应用。The VQG-23 dataset was developed by the Indian Institute of Technology Kharagpur for few-shot visual question generation tasks. It is expanded from the original VQG-16 dataset, with more image-question pairs and answer categories added, resulting in a total of 299,510 entries. During its development, the dataset was extracted from the Visual Genome dataset after removing duplicate images, ensuring data diversity and practical utility. The VQG-23 dataset is primarily used to evaluate and advance few-shot learning research, particularly for applications in vision-and-language generation tasks.