Project Overview
We
study the use of games and crowd sourcing to capture and analyze
scientific trajectory data. Advanced imaging tools give
scientists the ability to capture images from the microscopic scale to
the planetary scale. However, the scientific value of such images is often limited
by the manual and time consuming work required to process them.
Meanwhile, the public spends millions of hours playing
Farmville, Angry Birds and
Sudoku. This represents an opportunity to harness Human Power
Units (HPU) cycles for more productive but still
enjoyable work. Recently, it was shown that untrained citizen
scientists can be effectively used to help solve problems in various
scientific settings: from Galaxy Zoo, which helps astronomers with
image labeling, to FoldIt, which helps biologists with protein folding.
Our AngryAnts game employs citizen scientists to trace (i) static objects in an image and (ii) moving objects such as ants in a video. The goal is to tackle the challenging data validation problem of selecting an accurate trajectory for each object from the several (possibly inaccurate) trajectories submitted by the citizen scientists whom play the game. A simple quantitative consensus of several estimates is found by taking an average. However, computing the consensus of traces is not as straight forward, and we design algorithm which use clustering, geometric alignment, sampling, as well as other methods for this purpose.
Acknowlegements
Partial
support
from NSF grant 1053573 "ImageQuest: Citizens Advancing Biology with
Calibrated Imaging and Validated Analysis"