AffectVisDial
收藏资源简介:
AffectVisDial是一个大规模数据集,由阿卜杜拉国王科技大学创建,包含50,000个基于视觉的对话,每个对话包含10轮问答,总共有27,180个工作小时的数据。该数据集旨在研究情感解释和推理任务,特别是在视觉基础对话中情感的形成。数据集内容丰富,涉及多种视觉艺术作品,旨在通过对话捕捉参与者的情感反应和解释。创建过程采用实时通信协议,确保数据的质量和相关性。该数据集的应用领域包括开发情感感知AI系统,以及通过人类反馈增强模型的情感理解能力。
AffectVisDial is a large-scale dataset developed by King Abdullah University of Science and Technology. It contains 50,000 vision-grounded dialogues, each consisting of 10 rounds of question-answer exchanges, with a total of 27,180 working hours of collected data. This dataset is targeted at research on emotion interpretation and reasoning tasks, especially the emergence of emotions in vision-grounded dialogues. Featuring rich content covering a wide variety of visual artworks, it aims to capture the emotional responses and explanations of participants through dialogues. The dataset was constructed using real-time communication protocols to ensure the quality and relevance of the collected data. Its application domains include the development of emotion-aware AI systems, as well as enhancing models' emotional understanding capabilities via human feedback.

- 1Affective Visual Dialog: A Large-Scale Benchmark for Emotional Reasoning Based on Visually Grounded Conversations阿卜杜拉国王科技大学 · 2023年



