Autonomous visual model building based on image crawling through internet search engines

In this paper, we propose an autonomous learning scheme to automatically build visual semantic concept models from the output data of Internet search engines without any manual labeling work. First of all, images are gathered by crawling through the Internet using a search engine such as Google. Then, we model the search results as "Quasi-Positive Bags" in the Multiple-Instance Learning (MIL) framework. We call this generalized MIL (GMIL). We propose an algorithm called "Bag K-Means" to find the maximum Diverse Density (DD) without the existence of negative bags. A cost function is found as K-Means with special "Bag Distance". We also propose a solution called "Uncertain Labeling Density" (ULD) which describes the target density distribution of instances in the case of quasi-positive bags. A "Bag Fuzzy K-Means" is presented to get the maximum of ULD. By this generalized MIL with ULD, the model for a particular concept is learned from the crawled images of the Internet search engines. Experiments show that our algorithm can get correct models for the concepts we are interested in. Compared to the original Google Image Search, our algorithm shows improved accuracy.