In this paper, we describe how to build an incremental structured part model for object recognition. The proposed method explores both global structural information and multiple local features of objects for object model characterization. It use part models to represent structure nodes, which encode the local information of an object. The parts are learned through a segmentation and clustering process, and are used to form the part models in terms of multiple feature fusion and multi-class SVMs. The structured part model is then constructed by correlating different parts through a deformable configuration. Furthermore, we present an incremental learning strategy, which learns a part model by using only a small number of training samples. Annotated images with high entropies are used to update the trained model. The advantage of our method is that it captures the inherent connections of the semantic parts of objects and characterizes the structural ∗Corresponding author. Email addresses: firstname.lastname@example.org (Xiao Bai), email@example.com (Peng Ren), firstname.lastname@example.org (Huigang Zhang), email@example.com (Jun Zhou) Preprint submitted to Neurocomputing February 10, 2015 relationships between them. The proposed approach is evaluated on two datasets and demonstrates advantages over several state-of-the-art part-based methods in the literature.
Download Full PDF Version (Non-Commercial Use)