Mehmet Sinan Ascioglu

Eyes on You, Tarantino!

Visualization project seeking various directors' unconscious-art emerging from the eye movement patterns of spectators watching specific scenes from these movies.

Computers for the Rest of You,Spatial Design

Our focus while watching movies shifts continuously from side to side, actor to actress, dancer to assassin, making our eyes focusing continuously on the screen from left to right, top to corner, etc.. Sometimes in very fast jumps, sometimes in very smooth moves.. What we are not aware is that the smooth/complex patterns of these focus shifts of our eyes are different in each scene and movie and posses the signature and style of the director. "Eyes on You, Tarantino!" tries to visualize these patterns, and gives users the opportunity to explore the movie in reference to these patterns.
The patterns are obtained by tracking the eye movements of spectators with the eye tracking glasses, which has a camera attached to one of its lenses. Eye movements are recorded while spectators are watching the scenes, then this recordings are processed generating the eye movement data in xml format.
The visualization is projected on a column and a controller kiosk. Movies are displayed on the kiosk, on which the user can switch between the movies and browse back and forth within the displayed movie. The user is allowed to browse through the movie or through the eye movements pattern on which they can refer to the past/future of the eye movements and attached dialogs. The visualization algorithm gives more detailed view on the current time of the movie, but keeps the past visible, allowing users to see the big picture and compare the pattern to the patterns of other movies .

I spent a lot of time looking at the generative art made out of movies. Interesting pieces have influenced me to work with eye movements, which was not used as a direct input. A very nice example can be found here, in which the movements in the movie is put frame by frame, and itself become a tool for analyzing motion.

People interested in generative/algorithmic art, and film studies, movies.

User Scenario
When user approaches to the controller, the movie playing on the display (and the pattern projection flowing on the column) stops, waiting for user the explore it. User can go back and forth by pushing and pulling the controller up and down. User can switch between the movie scenes using the button on the controller.

The controller part which is attached to the column is made of semi-transparent blurry plexy. The actual movie either can be implemented below the plexy or projected from the projection above (this depends on the result of my experiment with the projection, keystone might be an issue). Rest is the column itself.
The eye movement pattern is projected perpendicular to the display, onto a large space (one of the columns at ITP) (see the prototype drawing and the videos at the project URL).
Project will show specific two minute scenes from Rear Window - Alfred Hitchcock, Death Proof - Tarantino, Tango - Carlos Saura, Hyperballad, Bjork Music Video - Michel Gaundry. These scenes spefically different from each other; Rear Window creates a square-wave like pattern in a dialog of two people, Death Proof creates a right to left shifting pattern in a monologue of a girl, Tango dance scene creates a pattern that uses all over the screen with a smooth pattern, and HyperBallad creates an amazing sinus wave pattern.

This project had been an amazing experience for me this semester, which I worked on pretty much for all of my assignments on Computers for the Rest of You class. It had multiple steps:
creating a good eye tracking software: in this part I was able to built my own eye tracking glasses, which brought a lot of challenges to tackle with. As my project end-product started shaping up, I decided to stay in horizontal detection.
Next phase was to visualize the information: At this stage, I was looking for a solution not only shows the interesting motion of the focus point of eye in a movie, but also which reveals the difference among movies, scenes, and directors from this perspective. So the visualization I am creating focuses on the 'current-time' eye movements but also shows the past, to give the user the big picture for that scene. So that, user will be able to compare two different scenes accordingly.