Dynamically Constructed Bayesian Networks for Sketch Understanding

People sketch to express their early design ideas in many domains, but current computer tools offer few advantages to designers during this sketching phase. Our goal is to construct a general recognition architecture that can be applied to a number of domains that is capable of parsing the user’s strokes (in real time) and interpreting them as depicting objects in the domain of interest without limiting the designer’s drawing freedom. Such an interpretation engine will enable the creation of powerful and natural early-stage computer aided design tools.