Reference is now made also to FIG. 4, showing a schematic illustration of an exemplary plurality of prediction models 400, according to some embodiments of the present invention. In such embodiments, the one or more prediction models are plurality of prediction models 400. Optionally, at least one of the one or more prediction models is a neural network. Optionally, the one or more prediction models comprise first prediction model 410A and second prediction model 410B. Optionally, first prediction model 410A is associated with a first set operator applied to a plurality of sets of labels. Optionally, the plurality of sets of labels is indicative of a plurality of features of one of the plurality of digital images. Some examples of a set operator are union, intersection and subtraction (set-difference). Optionally, second prediction model 410B is associated with a second set operator applied to the plurality of sets of labels. Optionally, in 310 processor 201 computes first intermediate group of features 420A by inputting into first prediction model 410A a first plurality of groups of features comprising group of features 401A and group of features 401A. Optionally, group of features 401A and group of features 401B are at least some of the plurality of input groups of features, for example group of features 111B and group of features 111C respectively. Optionally, processor 201 computes in 310 second intermediate group of features 420B by inputting into second prediction model 410B a second plurality of groups of features, optionally comprising second intermediate group of features 420A. Optionally, the second plurality of groups of features comprises at least some of the plurality of input groups of features, for example group of features 401C. Optionally, group of features 401C is group of features 111A. Optionally, processor 201 computes a third intermediate group of features 420C by inputting another group of features 401D of the plurality of input groups of features into third prediction model 410C, associated with another set operator applied to the plurality of sets of labels, where third prediction model 410C is one of the one or more prediction models. Optionally, processor 201 produces fourth intermediate group of features 430 by inputting third group of features 420C into fourth prediction model 410D, where fourth prediction model 410D is one of the one or more prediction models. Optionally, the output group of features is fourth intermediate group of features 430. Optionally, the output group of features is second intermediate group of features 420B. In other embodiments, the output group of features is first intermediate group of features 420A or third intermediate group of features 420C. Optionally, the first plurality of groups of features comprises two groups of features. Optionally, the first plurality of groups of features comprises more than two groups of features. Optionally the second plurality of groups of features comprises two other groups of features. Optionally, the second plurality of features comprises more than two other groups of features.