Communicating semantic part information in drawings

AbstractWe effortlessly grasp the correspondence between a drawing of an object and that physical object in the world, even when the drawing is far from realistic. How are visual object concepts organized such that we can both recognize these abstract correspondences and also flexibly exploit them when communicating them to others in a drawing? Here we consider the notion that the compositional nature of object concepts enables us to readily decompose both objects and drawings of objects into a common set of semantically meaningful parts. To investigate this, we collected data on the part information expressed in drawings by having participants densely annotate drawings of real-world objects. Our dataset contained both detailed and sparser drawings produced in different communicative contexts. We found that: (1) people are consistent in what they interpret individual strokes to represent; (2) single strokes tend to correspond to single parts, with strokes representing the same part often being clustered in time; and (3) both sparse and detailed drawings of the same object emphasize similar part information, although detailed drawings of different objects are more distinct from one another than sparse drawings. Taken together, our results support the notion that people flexibly deploy their abstract understanding of the compositional part structure of objects to communicate relevant information about them in context. More broadly, they highlight the importance of structured knowledge for understanding how pictorial representations convey meaning.


Return to previous page