A computer-implemented method and an apparatus for generating an action-to-feature mapping of a visual artwork may include initializing a system to create a visual artwork in, and obtaining an input stream of data recording a creation of a visual artwork from the system, and identifying, based on the input stream of data, at least one time series of utensil actions capable of reproducing the visual artwork, and obtaining at least one image of the visual artwork, and generating the action-to-feature mapping based on the at least one time series of utensil actions and on the at least one image of the visual artwork, thereby recognizing at least one artistic feature of the visual artwork. The action-to-feature mapping may include at least one pair of at least one utensil action of the at least one time series of utensil actions and the at least one artistic feature.