Special Issue on Generating Realistic Visual Data of Human Behavior

The fast and broad progress in AI has not only enabled great advances in the analysis of human behavior but has also opened new possibilities for generating realistic humanlike behavioral data. This special issue focuses on recent advances and novel methodologies for generating data containing human-like behaviour atmultiple scales, and contains eight papers, summarily described. Two types of applications, generation of realistic facial behaviours and body movements, are investigated and represented by four papers each.Methodologically, generation of realistic synthetic data helps with automatically established ground truth labeling and data augmentation, as well as for animation and content generation. The contributions from all of these perspectives are presented in this special issue, which brings together eight contributions, selected from 17 submissions following an open call for papers. Each paper was rigorously peerreviewed for one or two rounds of revisions according to the journal’s high standards. The first paper in the series, “Generating Human Action Videos by Coupling 3D Game Engines and Probabilistic Graphical Models” by De Souza, Gaidon, Murray, Cabon, and López is on generation of synthetic training data for video based action recognition. With an interpretable parametric model that combines elements of game engines, the proposed approach generates videos for natural and parametrically defined action sequences. These videos are easily employed for training data augmentation, boosting the recognition performance of classifiers they are used with. The