DOLOS
收藏资源简介:
DOLOS数据集是南洋理工大学创建的音频-视觉欺骗检测数据集,包含1675个视频片段,涉及213个对象。该数据集源自英国的现实喜剧游戏节目,每个视频片段都标注了音频-视觉特征。DOLOS数据集旨在通过提供丰富的欺骗对话样本,推动多模态欺骗检测的研究。数据集不仅用于欺骗检测,还可应用于面部表情预测等多任务学习,为研究者提供了一个高质量的研究平台。
The DOLOS Dataset is an audio-visual spoofing detection dataset developed by Nanyang Technological University. It contains 1675 video clips involving 213 subjects. This dataset originates from a British reality comedy game show, and each video clip is annotated with audio-visual features. The DOLOS Dataset aims to advance research in multimodal spoofing detection by providing abundant spoofed dialogue samples. Apart from spoofing detection, the dataset can also be applied to multi-task learning tasks such as facial expression prediction, serving as a high-quality research platform for researchers.

- 1Audio-Visual Deception Detection: DOLOS Dataset and Parameter-Efficient Crossmodal Learning南洋理工大学 · 2023年



