kd7/graid-bdd100k

GRAID BDD100K问答数据集是一个基于BDD100K数据集生成的问答对集合，主要关注空间推理、物体定位和视觉推理。它包含各种类型的问题，旨在测试对图像的不同理解方面。数据集分为训练集和验证集，并包含图像、注释、问题、答案、问题类型和源ID等特征。该数据集使用GRAID框架生成，该框架是一种从物体检测注释中创建问答对的方法。数据集根据BSD-3-Clause许可证提供，并通过HuggingFace数据集库可供下载。

The GRAID BDD100K Question-Answer Dataset is a collection of question-answer pairs derived from the BDD100K dataset, focusing on spatial reasoning, object localization, and visual reasoning. It includes various types of questions designed to test different aspects of understanding within images. The dataset is split into training and validation sets, and it contains features such as images, annotations, questions, answers, question types, and source IDs. The dataset is generated using the GRAID framework, which is a method for creating question-answer pairs from object detection annotations. The dataset is licensed under the BSD-3-Clause license and is available for download through the HuggingFace datasets library.