Children often portray their feelings through drawings, in spite of being oblivious to incorrect adult behavior. Drawings often are a gate to children feelings which can’t yet be communicated through words, but which we can interpret through drawing. A study in 2014, for instance, identified a large gap between abused children victims who drew and those that were not given the option .
An interesting use case would be using a feature space of drawings to test if it is able to identify drawings from abused children from those that were not. The choice of colors, recognizable shapes, and actions could be a starting point for feature engineering.
Drawings may contain in numerous details when interpreted. Some easy to spot, and some not. Neural Networks, and even more so Deep Learning shares a graph model where one can represent feature dependency through an underlying construct. The interpretation of a children’s mind through drawing that was victims will be sharing a common theme: Abuse.