FE-3DGQA
收藏arXiv2022-09-24 更新2024-07-31 收录
下载链接：
http://github.com/zlccccc/3DGQA
下载链接
链接失效反馈官方服务：
资源简介：
FE-3DGQA是一个新收集的3D视觉问答数据集，由北京航空航天大学软件学院创建。该数据集包含20,215个自由形式的问答对，以及密集和完全基于地面的边界框注释，用于703个室内场景。数据集的创建过程涉及手动注释，确保了问题和答案的多样性和复杂性。FE-3DGQA数据集的应用领域主要集中在3D视觉问答任务，旨在通过提供详细的视觉和语言信息，帮助模型理解和回答关于3D场景的复杂问题。
FE-3DGQA is a newly collected 3D visual question answering dataset developed by the School of Software, Beihang University. This dataset contains 20,215 free-form question-answer pairs, as well as dense and fully grounded bounding box annotations for 703 indoor scenes. The dataset construction process adopts manual annotation, which guarantees the diversity and complexity of the questions and answers. The application scope of the FE-3DGQA dataset mainly focuses on 3D visual question answering tasks, aiming to help models understand and answer complex questions about 3D scenes by providing detailed visual and linguistic information.
提供机构：
北京航空航天大学软件学院
创建时间：
2022-09-24
