The DIEM Project (Dynamic Images and Eye Movements) uploaded a pair of 6:20 clips from There Will be Blood, which includes location pointers based on the viewing habits of eleven people watching a particular scene from the film. Why do this? Well, according to the org’s Vimeo page, “Because human visual perception involves active information seeking via eye movements, much of the work in the lab focuses on eye movements and human gaze control.”
Beyond the curiosity factor I’m sure directors and cinematographers could use this information to help them frame their shots, though there wasn’t much that I saw during the clip’s duration that seemed all that out of the ordinary or surprising. However, after the first video, we have a second look at the same scene except instead of seeing the full image, all that’s visible are the elements of the scene with the highest concentration of visual attention.
Probably what I found most fascinating about these videos is the perceived curiosity by viewers to the onlookers and the reactions of not the people in direct conversation, but those in the background and occasionally slightly blurry. Slight background movement seems to register almost every time.
This is an excerpt from There Will Be Blood (Paul Thomas Anderson, 2007). 11 adult viewers were shown the video and their eye movements recorded using an Eyelink 1000 (SR Research) infra-red camera-based eyetracker. Each dot represents the center of one viewer’s gaze. The size of each dot represents the length of time they have held fixation.
A clip from There Will Be Blood presented as only the locations fixated by 11 viewers. Their gaze was recorded using an Eyelink 1000 and visualised as a “peekthrough” heatmap using CARPE
Now, what I found to be more interesting, simply because it is a faster moving example, is the following use of the same technology while viewers watched a trailer for Harry Potter and the Half-Blood Prince. It’s fascinating to watch as the eyes bounce from one target to the next in the quickly edited piece and I can’t help but now wonder what it would be like if the dots were color-coded blue for male and red for female or any other telling manner for splicing the data.
To check out all of the DIEM Project’s videos, and there are several of them, click here.
