ShenzhenFace-EEG Dataset
The dataset, ShenzhenFace-EEG, is prepared for research on brain-perceived face recognitions, where 2500 face images are collected and selected from the widely researched database LFW [1] and FaceScrub [2]. Among the 2500 face images, 50 subjects can be identified, each of which contains 50 facial images. To facilitate the dataset collection and preparation, we selected 50 reference face images for human recognitions, providing a benchmark for our computerized brain activity analysis and face recognition. To enable an objective ground truth for the brain-perceived face recognition research, we select the 50 reference images upon two conditions: (i) faces are located in the middle, closest possible to passport standard among all 50 available images for each subject; (ii) maintains the highest possible resolution without any occlusion or ambiguity. As a result, all face images can be represented as follows:
Where each row represents one subject with 50 images, and the last one R is the reference image allocated for human recognition.
Stimulation is arranged as such each subject views every face image for 1200 ms, where the first 100ms and the last 100ms is removed when EEGs are collected to minimize the possible interferences.
References:
[1] http://vis-www.cs.umass.edu/lfw/
[2] http://vintage.winklerbros.net/facescrub.html
