FOWS (Face Occlusion With Swapping)

该数据集名为FOWS，包含了展示人类主体进行导致面部遮挡的自愿动作的视频，其中既有真实的视频片段，也有经过处理的视频片段。该数据集还包括通过面部交换算法（SimSwap、GHOST和FaceDancer）生成的三种处理版本，涉及7名志愿者执行六种挑战变体。规模上，数据集共有168个视频（42个真实视频和126个处理视频），大约70,000个视频帧。其任务是检测视频帧中与遮挡相关的面部交换伪迹。

This dataset is named FOWS. It contains videos capturing human subjects performing voluntary actions that induce facial occlusion, including both authentic video clips and manipulated video segments. Additionally, the dataset provides three manipulated versions generated via three face-swapping algorithms: SimSwap, GHOST, and FaceDancer, which involve 7 volunteers executing six challenge variants. In total, the dataset consists of 168 videos (42 authentic videos and 126 manipulated videos), amounting to approximately 70,000 video frames. The corresponding task for this dataset is to detect occlusion-related face-swapping artifacts in video frames.