worldcuisines/vqa

WorldCuisines是一个大规模的视觉问答（VQA）基准数据集，旨在通过全球美食进行多语言和多文化理解。该数据集包含30种语言和方言的文本-图像对，涵盖9种语言家族，包含超过100万个数据点，是截至2024年10月17日最大的多文化VQA基准。数据集包括两个主要部分：WC-VQA（视觉问答数据集）和WC-KB（世界美食知识库）。WC-VQA数据集基于WC-KB构建，包含两个任务：任务1是菜品名称预测，任务2是位置预测。数据集通过Wikipedia和Wikimedia Commons收集，并经过严格的元数据标注和质量保证过程。

WorldCuisines is a massive-scale visual question answering (VQA) benchmark for multilingual and multicultural understanding through global cuisines. The dataset contains text-image pairs across 30 languages and dialects, spanning 9 language families and featuring over 1 million data points, making it the largest multicultural VQA benchmark as of 17 October 2024. The dataset includes two main tasks: dish name prediction and location prediction. The construction process of the dataset includes dish selection, metadata annotation, quality assurance, and data compilation. The data sources include Wikipedia and Wikimedia Commons, ensuring that the data can be redistributed under an open-source license. The dataset also includes a knowledge base (WC-KB), containing 2,414 global dishes with 6,045 images and metadata, covering coarse-grained and fine-grained categories, locations, and regional cuisine information. The generation process of VQA data includes similarity search for dish names, construction of questions and contexts, multilingual translation, and generation of VQA triplets.