In the recent years, researchers have investigated several methods of facial expression analysis. Their interest has been to apply their algorithms to sets of images labelled by a restrained number of experts. In order to deal with the generalizability of the proposed methods, databases of numerous facial expressions images have been collected. Less attention has been given to the experts. In this paper we present a web-based survey aiming at collecting the judgment on different facial expressions of a heterogeneous number of experts. In order to provide the researchers with a common set of features for a fair comparison of algorithms, the collected data combine the participants choices and a set of measures computed on the images. The resulting database consists, at the moment, of 40704 annotations from 1785 experts.