The ability to monitor and record human behavior has increased exponentially in recent time, and while this seems like a boon for everyone interested in studying or monitoring this behavior, the ability to process the data is not keeping up with the volumes being collected. This is especially true in cases where the data is not easily analyzed by computers, but requires human involvement or oversight, such as face- or object-recognition tasks, image characterizations, or complex comprehension tasks.
Fortunately, a new technology is emerging that may be able to assist in the processing of these huge amounts of data. This has been alternatively termed "peer production" or "crowdsourcing" and amounts to divvying up a task into small parts that are easy for humans to do. Rather than asking one person to labor over a menial, repetitious task, the small parts are farmed out a to a large number of workers, each of whom only completes a small part, but the end result is a massive amount of data processing.
To test the feasibility of these peer production tools, we are using Amazon's Mechanical Turk to see the efficacy for a variety of different tasks. Presently, we are interested in whether people can take a randomly ordered set of snapshots taken from traffic camera and put them into chronological order. Other tasks we are considering include identifying a particular vehicle from these traffic images, identifying the number of people, vehicles, or buildings in different surveillance images, and categorizing the speed and density of the traffic.
Ideas for other tasks are very welcome, as are comments on the project as it stands.
These days people are so caught up in their everyday lives that they become selfish in a sense and fail to notice or pay even a slight bit of attention to what goes on around them. I would be interested to see (in a large group) how many acually cared enough to pay close enough attention in such memory games!!!
Comment by Amber (November 3, 2008, 11:11 am)Tell us what you think, leave a comment: