Ph.D. Candidate in Mathematical and Computational Cognitive Science

Department of Psychological Sciences

Purdue University

703 Third Street

West Lafayette, IN 47907

Email: li135@purdue.edu

Tel: (765)494-6865

My research focuses on exploring human perception from the point of view of psychophysics, mathematics and computer science. In my current project I am studying the mechanisms of human 3D shape perception and simulating these mechanisms with computational models.

When a 2D retinal image is formed by a 3D object, the depth information is lost. Therefore, producing a 3D shape percept is equivalent to recovering the 3D shape from the 2D retinal image. I proposed that the essence of the 3D shape percept is the operation of simplicity constraints, such as symmetry, planarity. The constraints allow the visual system to identify (select) the simplest 3D interpretation from among infinitely many possible 3D interpretations of the 2D image. Based on this idea, I developed a computational model that can recover 3D shapes from a single 2D image (a patent application is pending). The 3D shape recovery produced by this model is very close to the human percept. The shape recovery is quite stable in the presence of changes of the 3D viewing direction and in the presence of noise. This model can recover a wide range of 3D shapes, including shapes of real objects, such as furniture, animals, and cars. Please refer to the demo for examples of the models' recovery.