DOI

10.17077/etd.pmcjyoy5

Document Type

Thesis

Date of Degree

Spring 2017

Access Restrictions

.

Degree Name

MS (Master of Science)

Degree In

Industrial Engineering

First Advisor

Baek, Stephen

First Committee Member

Baek, Stephen

Second Committee Member

Lendasse, Amaury

Third Committee Member

Song, Xuan

Abstract

Game with a purpose (GWAP) is a concept that aims to utilize the hours spent in the world playing video games by everyday people to yield valuable data. The main objective of this research is to prove the feasibility of using the concept of GWAP for the segmentation and labeling of massive amount of 3D human body scan data. The rationale behind using GWAP as a method for mesh segmentation and labeling is that the current methods use expensive, time consuming computational algorithms to accomplish this task. Furthermore, the computer algorithms are not as detailed and specific as what natural human ability can achieve in segmentation tasks. The method presented in this paper overcomes the shortcomings of computer algorithms by introducing the concept of GWAP for human model segmentation. The actual process of segmenting and labeling the mesh becomes a form of entertainment rather than a tedious process, from which segmentation data is produced as a bi-product. In addition, the natural capabilities of the human visual processing systems are harnessed to identify and label various parts of the 3D human body shape, which in turn gives more details and specificity in segmentation. The effectiveness of the proposed game play mechanism is proven by experiments conducted in this study.

Public Abstract

Games with a purpose (GWAP) is a concept that aims to utilize the hours spent in the world playing video games by everyday people to yield valuable data. The main objective of this research is to prove the feasibility of using the concept of GWAP for the segmentation and labeling of massive amount of 3D human body scan data. Accurate segmentation and labeling of 3D human body scan data is of paramount importance in many areas of research, such as human tracking, digital human modeling and anthropometric analysis, and etc. Such a task is extremely expensive as it requires lots of tedious, time-consuming hand labor. There are a number of computational methods for automatically segmenting and labeling a 3D model, but the computers are still a way less reliable than the human capability in terms of accuracy, amount of detail, and robustness. To solve this problem and make geometry labeling and segmentation easier and less demanding, we created a multiplayer online game called “Mosquito Popper”. The purpose of this game is to re-channel the efforts and the hours spent by game players to the collection of segmentation and labeling data through gameplay. By crowdsourcing, it is possible to yield massive amounts of segmented mesh data while at the same time accurately label the 3D human body mesh segments. The advantage this approach offers is that it will help overcoming the problem of tedious and time consuming work that previous computational approaches ran into. While playing the game, the players will effectively be providing geometry labeling and mesh segmentation data through the mechanics of the game. Furthermore, by collecting this data directly from the players this way, we are using the natural 3D segmentation and image identification capability that humans already possess. Finally, we ultimately want to create a data repository full of already segmented and labeled 3D human body scan data that is easy to navigate and use for future research.

Keywords

Games With a Purpose (GWAP), Human Body Scan, Human Computation, Labeling, Segmentation

Pages

ix, 35 pages

Bibliography

Includes bibliographical references (page 35).

Copyright

Copyright © 2017 Zachary Nolte

Download

Share

COinS