MovieFIB

MovieFIB是一个基于描述性视频注释的填空式问答数据集，专为视觉障碍人士设计，包含超过300,000个问题-答案和视频对。数据集由蒙特利尔大学创建，旨在通过填空问题评估视频内容的理解。数据集内容包括实体、动作和属性，通过自动生成填空问题，无需额外手动工作即可扩展至大量查询。该数据集的应用领域包括内容推荐、预测、摘要、事件/对象检测及理解人类视觉感知，旨在解决视频理解中的量化基准需求。

MovieFIB is a fill-in-the-blank question answering dataset built on descriptive video annotations, specially designed for visually impaired individuals. It contains over 300,000 question-answer and video pairs. Developed by the University of Montreal, this dataset is intended to assess video content comprehension via fill-in-the-blank questions. The dataset covers entities, actions and attributes, and automatically generates fill-in-the-blank questions, enabling expansion to a massive volume of queries without requiring extra manual work. Its application domains include content recommendation, prediction, summarization, event/object detection, and the investigation of human visual perception, with the goal of addressing the demand for quantitative benchmarks in video understanding.