Read this if you want to know more about what I have done so far and what I like in terms of research.

My work as research assistant under Dr. Elena Fedorovskaya was to come up with ways to understand Human Perception of Art, specifically digital images of paintings. To be more specific, the goal was to see if there are any differences in viewing patterns of experts and novices while viewing paintings. As every other research work, I read a lot of research papers in this area. My work then began with collecting dataset. As part of this, I was introduced to Eye-Trackers. Both remote (one that is fixed to a desktop) and a wearable Eye-Tracker. This work led to a publication that you can read here

Because I liked what I was doing, I chose to continue this for my master thesis. This work was co-advised by Dr. Christopher Kanan . As part of my thesis, I first implemented the Multi-Fixation Pattern Analysis (MFPA) from the paper on the data we collected. MFPA is a family of machine learning algorithms for making inferences about people from their scanpaths. Little briefly about this algorithm - Fixation location (x,y) is taken and converted to Fisher Vectors before feeding them into a classifier for classification. It has been shown that if MFPA algorithms can do an inference above chance, this suggests that there are differences between scanpaths for the variable of interest. But this did not work on our dataset. So we came up with an idea that not only modified the existing algorithm but also gave us concrete results that showed that there exists differences between our experts and novices.

Our Idea and Approach.

Instead of turning the fixation locations into Fisher Vectors, We extracted image features surrounding these locations using a deep Convolutional Neural Network and converted these features into Fisher Vectors before feeding them into a classifier. We extracted both low- and high-level image features to see if there are differences in viewing patterns of our experts and novices when both bottom-up and top-down factors were taken into consideration. Although we cannot still confirm on who is contributing more to what features, we can confirm that there is definitely a difference between experts and novices under all cases we took into consideration. This work is currently under review at VISARTIV . But if you're interested to know more do not hesitate to contact me.


In my opinion, research is like addiction. My first encounter with research was during my undergrad. What happens in brain of somebody who's diagnosed with Alzhiemers? Can I study it with Brain imaging? These naive questions for which I did not have answers led me to RITs Imaging Science program. You read something and you want to know more. And with my graduate education my questions only kept increasing. Over the period of my time here, I developed a habit of reading research papers. Many times, I did not understand the deep technical details of the papers, but still I would read because curiosity would kill me and I would keep questioning why's and what's and where's and why this professor or researcher is doing this and where all their work is leading to. Viola! This only got me interested in the intersection of Geometry, Cognitive Science and Vision in general. I wish I could just study everything about brain, vision, mathematics and computers :)