Faces in Event Streams (FES): An Annotated Face Dataset for Event Cameras

The dataset contains 689 minutes of recorded event streams and over 1.6 million annotated faces with bounding boxes and five point facial landmarks marked at a 30Hz rate. Event streams were recorded on Prophesee PPS3MVCD event camera with a resolution of 480×360, 30 μm pitch and FOV of 70 degrees. Dataset consists of two parts based on the controlled and uncontrolled manner of data collection and difference in number of participants, overall featuring 73 unique subjects.该数据集包含时长689分钟的录制事件流，以及超过160万张带有边界框与五点面部标记点的标注人脸，数据采集帧率为30Hz。事件流由分辨率为480×360、像素间距30μm、视场角(FOV)为70度的Prophesee PPS3MVCD事件相机录制。该数据集依据数据采集的受控与非受控模式，以及参与人数的差异分为两个子集，总计涵盖73名独特受试者。