Exploring human computation via interactive online application

• Main Author: Lee, Ming Hua.
• Other Authors: School of Computer Engineering
• Format: Final Year Project
• Language: English
• Published: 2010
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: http://hdl.handle.net/10356/36262
• Institution: Nanyang Technological University

Training a computer vision algorithm to test whether an image contains a particular object, a dog for example, would involve presenting it with multiple images of a dog, each annotated with the precise location of the dog in the image. After processing enough images, the algorithm learns to find dogs in arbitrary images. A major problem with this approach, however, is the lack of training data which must be prepared by hand. This is a FYP Final Report to document all relevant background information, the game specification, its test results and findings regarding the project "Exploring Human Computation via Interactive Online Application". Tag! is a Game With A Purpose (GWAP); the interactive online application that make use of Human Computation to address the problem of collecting large amount of training data for computer vision algorithm. It has the purpose of improving image search and indexing as well. Tag! is an online game application where players are randomly grouped in threes and placed in a room to compete round after round. As simple as the game title suggests, players will have to tag as many objects in that image as possible in each round. Subsequently, these tags will help to segment an arbitrary image into appropriate sub-images, each depicting an object with meaningful labels. In February 2010, Tag! has its first playable version completed and is put for a closed beta trial test. This version consists only of the basic features required for a proper game play. The results are stored and analyzed, and turn out to be not quite expected.