aist-cvrt/HanDyVQA

HanDyVQA数据集是一个用于评估手-物体交互中时空动态、过程和效果的新基准数据集。它基于Ego4D数据集构建，包含视频片段和对应的问题与答案。数据集分为训练集、验证集和测试集，每个数据集都以JSONL格式存储，每行是一个JSON对象，包含问题的类型、ID、视频ID、问题描述、答案选项、正确答案索引等信息。

The HanDyVQA Dataset is a new benchmark for evaluating abundant spatio-temporal dynamics, process, and effects in hand-object interactions. It is built on top of the Ego4D Dataset and contains video clips along with corresponding questions and answers. The dataset is split into training, validation, and test sets, each stored in JSONL format, with each line being a JSON object that includes the question type, ID, video ID, question text, answer options, indices of the correct answers, and more.