Human computation and labeling image wtih computer games

Despite the recent advances in software, hardware and related techniques, no satisfactory solution to date has been developed for fully automated image labeling problem. Hence, to ease the problem, the recent work in the literature studied semi-automated approaches which mostly depend on leveraging the computational power of human brain. Human computation is one of such approaches, and based on the principle of injecting hard (for computers) problems into computer games. Clearly, while entertaining with such games, people are essentially contributing to the solution of the respective hard problem. There are several computer games developed at solving image labeling problem, including Peekaboom, Phetch, Verbosity and ESP game. In the context of this thesis, yet another human computation based image labeling system is designed and implemented. Since the computation is human-centric, image labeling differs from culture to culture. For this reason, the system is used to label images belonging to the local culture. The work also presents the performance evaluation of the system.