berhaan/clevr-tr

LLaVA-CoT数据集是一个用于视觉问答（VQA）任务的开源数据集，支持英语和土耳其语。数据集包含问题和图像，答案由GPT-4生成。数据集的结构包括一个`train.jsonl`文件，其中包含问答数据，格式为JSON，包含图像路径、对话内容及其英文翻译。数据集的创建基于CLEVR和CLEVR-Math数据集，分别用于一般VQA和科学目标VQA任务。数据集的规模为小于1K。此外，数据集可能存在一些偏见和风险，用户可以通过提供的联系信息处理侵权或不适当内容。

The LLaVA-CoT dataset is an open-source dataset for visual question answering tasks, containing English and Turkish languages. The dataset structure includes a train.jsonl file, which contains question and answer data, along with corresponding image paths. The dataset creation process used open-source datasets and generated answers. The dataset may have certain biases and risks, but efforts have been made to ensure data diversity and accuracy.