Improving Sculpting via Crowdsourcing

Abstract

Crowdsourcing and the wisdom of crowds [5] have established that a large number of low quality inputs can be reliably aggregated to yield a higher quality output. This thesis examines whether the wisdom of crowds can be extended to sculptures or three-dimensional models. To do so, we discuss the design of a 3D modeling tool requiring no training and make use of it to crowdsource the creation of hundreds of sculptures from nearly a thousand individuals. We introduce several techniques for aggregating low quality 3D models. Finally, we performed a perceptual study to evaluate the quality of input and aggregate models. We report our ndings related to additive versus subtractive modeling, automatic versus manual rotation, and our various aggregation techniques.