Abstract: Typical photo-based augmented reality applications match photos against contents in a database of images, using image retrieval algorithms. Further, in such augmented reality applications, content authoring is a tedious and expensive task that requires specialized skills. Consequently, scaling database contents is hard, which results in lack of data sets to be recognized and even reduces accuracy in recognizing objects. We propose a mobile augmented reality software framework that enhances scalability of contents by using human computation resources. We utilize social network services as resources for our system, which collect image content generated by active users in social network services such as Twitter. As a result, we provide very large and scalable database of contents. We demonstrate that our new system is not only feasible but a practical solution.