I am a member of the Design Rationale Group in CSAIL. My main research goal is enabling people to interact with computers in a more natural fashion through amalgamation of techniques from computer vision, machine learning, computer graphics and human computer interaction. Specifically, I am interested in developing intelligent pen based human-computer interfaces that can understand simple free-hand drawings.

My PhD thesis explores ways of using hidden Markov models, and dynamic Bayesian networks to model sketching as a stochastic dynamic process such that we can learn models of stroke ordering regularities present in online sketching, and use these models to do sketch recognition.

Sketch recognition is hard, in part because sketches are messy and highly variable in a way that is unlike images. Part of my PhD work also involves observing professionals that use sketching in their work, collecting real data from these professionals and analyzing the data. I focus on electrical circuit diagrams produced by engineers, and course of action diagrams and map annotations used by army personel -- for which I got to visit the US Army National Training Center in California.

In my master's thesis, I explored multi-scale shape approximation techniques. In particular I combined ideas from the scale-space literature with a novel scale selection algorithm to achieve feature point detection and shape approximation in digital curves.

In the past, I have spent time experimenting and learning about other topics including numerical methods, simulation, dynamical systems, and chaotic systems. Thanks to my father Fatin Sezgin, I have also developed an interest in random number generators. Being able to write a Gibbs sampler from scratch is always fun.

If you are interested to learn more about my research, check out my papers.