hoangphann/ViFoodVQA

ViFoodVQA是一个越南食品知识视觉问答基准数据集。每个样本包含一张越南美食图片、一个基于食品知识图谱的越南多选题、四个选项、正确答案、答案解释以及用于基础的KG三元组。数据集结构包含9个字段：vqa_id（唯一标识符）、image_id（图片标识符）、image（图片路径）、qtype（问题类型）、question（问题文本）、choices（选项字典）、answer（正确答案）、rationale（答案解释）和triples_used（KG三元组列表）。问题类型涵盖食材、食材类别、烹饪技术、食物搭配、饮食限制、过敏原限制、菜肴分类、风味特征、原产地和替代规则等。数据集分为训练集、验证集和测试集，其中测试集经过人工验证。

ViFoodVQA is a Vietnamese food knowledge-based visual question answering benchmark. Each sample contains an image of Vietnamese cuisine, a Vietnamese multiple-choice question grounded in a food knowledge graph, four answer choices, the correct answer, a rationale explaining the answer, and the KG triples used for grounding. The dataset structure includes 9 fields: vqa_id (unique identifier), image_id (image identifier), image (image path), qtype (question type), question (question text), choices (answer choices dictionary), answer (correct answer), rationale (explanation for the answer), and triples_used (list of KG triples). Question types cover ingredients, ingredient categories, cooking techniques, food pairings, dietary restrictions, allergen restrictions, dish classification, flavor profiles, origin locality, and substitution rules. The dataset is split into train, validation, and test sets, with the test set being human-verified.