AJN-AI/VoQA

VoQA（视觉问答）数据集是一个专注于仅使用视觉信息进行问答任务的数据集。它包含了训练集和评估集，评估集进一步细分为多个子任务，如GQA、POPE、ScienceQA、TextVQA和VQAv2。每个子任务都有特定的数据格式。训练集包含coco、gqa、ocr_vqa、textvqa、vg/VG_100K和vg/VG_100K_2等任务。README文件中提供了下载和设置数据集的详细说明以及其原始结构。

The VoQA (Visual Question Answering) dataset is focused on question answering tasks using only visual information. It consists of a training set and a benchmark (test) set, with the benchmark further divided into sub-tasks such as GQA, POPE, ScienceQA, TextVQA, and VQAv2. Each sub-task has specific data formats. The training set includes tasks like coco, gqa, ocr_vqa, textvqa, vg/VG_100K, and vg/VG_100K_2. The README provides detailed instructions for downloading and setting up the dataset, as well as its original structure.