Figure 1

The action concepts dataset. (a) Example footages (from the publicly available dataset¹⁷) related to some of the selected actions for the task of concept understanding, i.e., , recognizing actions that are implied as opposed to be explicitly visualized. The “Cooking” action is represented by the smoke coming out from a pan, the creation of a dough, the mixing of a chocolate cream, or the garnishing of a dessert (top-left). The “Flying” action is represented from footages in which pigs fly, fireworks are shot in the sky; alternatively, it is also depicted as a scene in which the panorama is filmed from an airplane window, and as Santa Claus delivering presents (top-right). For “Hugging”, unusual cases are considered such as a man hugging his dog, a girl hugging a tree, two gibbons hugging each others, and a baby hugging his toy (bottom-left). Similarly, for “Throwing”, elephants can throw sand on their backs, axes can be thrown, a weight can be thrown during a fitness sessions, and eventually money can be thrown in air (bottom-right). (b) Comparison of existing public benchmarks for EEG data processing, finalized to several applications.