kurt

Morning Midterm Thoughts

While discussing the midterm project that Tripp and I are collaborating on, my concerns centered around the capability of current technology to analyze video in a way that is interesting to a user that hopes to discover something about a "character" by viewing a continuous POV shot. The hardware/software that we will use will be off the shelf, meaning essentially a webcam taped to a head and Max/MSP Jitter or SoftVNS to evaluate/manipulate the video. Once a user hands the footage over to the system, there are severe limits on the kind and quality of metadata that can be harvested. I don't mean to underestimate the value of a digital sort on video content, I am most worried that such a mechanized look at human perspective will tell us more about the system's capabilities than about the internal mental state of the user.

So I propose that we simultaneously pursue a minimally intrusive "live" user annotation system that would allow points of interest to be marked while in the field. By merely adding a flag at specific points in the timecode, we can learn about the moments that are meaningful to the user. I know that Tripp is primarily interested in what a data management system can accomplish without any hints from the user, but by allowing ourselves the insight into the sequences that the user cares about, we can then isolate those shots and try to figure out what they have in common.

Ultimately I am interested in a model of active capture vs. passive capture for expression, so the idea of recording and storing everything that I see is initially content-less to me. There is immense value however in the capability to record and store everything, a capability that we are inevitably approaching, especially when the user can match his/her thoughts to an image or a sound or any other sensed media.