VDialogUE

VDialogUE是由深圳先进技术研究院和中国科学院联合创建的视觉基础对话评估基准，旨在通过统一的标准评估视觉基础对话系统。该数据集包含六个不同的数据集，覆盖五个核心多模态对话任务，包括多模态意图预测、多模态对话检索（文本到图像和图像到文本）、多模态对话状态跟踪和多模态响应生成。VDialogUE的创建过程涉及精心设计和选择数据集，以确保评估的全面性和准确性。该数据集的应用领域广泛，旨在解决视觉基础对话系统在多模态环境下的性能评估问题，推动该领域的技术发展和模型优化。

VDialogUE is a visual grounded dialogue evaluation benchmark jointly developed by the Shenzhen Institute of Advanced Technology and the Chinese Academy of Sciences, aiming to evaluate visual grounded dialogue systems via unified standards. This dataset comprises six distinct datasets covering five core multimodal dialogue tasks, including multimodal intent prediction, multimodal dialogue retrieval (text-to-image and image-to-text), multimodal dialogue state tracking, and multimodal response generation. The creation of VDialogUE involves meticulous design and selection of datasets to ensure the comprehensiveness and accuracy of the evaluation. This benchmark has wide-ranging applications, aiming to address the performance evaluation challenges of visual grounded dialogue systems in multimodal environments, and promote technological development and model optimization in this field.