Image-Chat

该数据集是由众包工作者收集的高质量图像基础开放领域对话，对话内容基于给定的图像和两种个性特征展开。在进行研究时，这些对话中的个性信息被忽略。该数据集的规模包括186,782条训练数据，5,000条验证数据以及9,997条测试数据，其任务是图像基础对话。

This dataset is a collection of high-quality image-grounded open-domain conversations collected by crowdworkers. Each conversation is developed based on a given image and two personality traits. During the research, the personality information within these conversations was omitted. In terms of scale, this dataset contains 186,782 training samples, 5,000 validation samples, and 9,997 test samples, with the targeted task being image-grounded conversation.