sparks-solutions/AutomotiveUI-Bench-4K

AutomotiveUI-Bench-4K是一个专注于车内信息娱乐系统（IVI）交互的视觉问答数据集，包含998张图像和4208个注释。数据集涵盖了15个汽车品牌/OEM，车型年份从2018年到2025年。主要特征包括图像、边界框、类别标签以及针对测试动作和预期结果的文本描述。注释语言为英语，部分德语文本被翻译或引用。该数据集被用作汽车用户界面的验证基准，并可与Hugging Face上的ELAM模型兼容。

AutomotiveUI-Bench-4K is a visual question answering dataset focusing on interactions with in-vehicle infotainment (IVI) systems, containing 998 images and 4,208 annotations. It covers 15 automotive brands/OEMs, model years 2018-2025. Key features include images, bounding boxes, class labels, and textual descriptions for test actions and expected results. Annotations are in English, with some German UI text translated or quoted. The dataset serves as a validation benchmark for automotive UI and is compatible with the ELAM model available on Hugging Face.