标题：Unsupervised Object Localization with Latent Dirichlet Allocation
作者：Yang, Tong-feng; Ma, Jun
作者机构：[Yang, Tong-feng; Ma, Jun] Shandong Univ, Sch Comp Sci & Technol, Jinan 250100, Peoples R China.
会议名称：International Conference on Computer Science and Artificial Intelligence (ICCSAI)
会议日期：NOV 16-17, 2013
来源：2013 INTERNATIONAL CONFERENCE ON COMPUTER SCIENCE AND ARTIFICIAL INTELLIGENCE (ICCSAI 2013)
关键词：object localization; object detection; latent dirichlet allocation
摘要：Current supervised learning approaches to object localization require datasets of training images to be manually prepared, with varying degrees of supervision. We conclude three heuristic rules for unsupervised object localization and proposed a method under the rules: We use Gaussian Mixture Model to characterize the position and shape of objects, and modify the form of LDA to provide location and scale estimation of the foreground object. We propose a Gibbs Sampling algorithm to estimate the parameters of the modified LDA as well. In experiment, we show results with significant improving over several published algorithms.