Mining Frequent Feature Configurations for Architecture Image Classification and Product Image Search

Wei-Ta Chu and Ming-Hung Tsai

Multimedia Computing Laboratory
Dept. of Computer Science and Information Engineering
National Chung Cheng University

1. Introduction

Repetitive elements, or patterns, are ubiquitously presented in man-made objects and natural environments. Many objects have characteristic repetitive elements, hence finding repetitive patterns facilitates object recognition, image analysis, and other applications. We devise a novel description, i.e. visual pattern, to describe feature configurations of repetitive elements, and present an approach to automatically detect and localize visual patterns in images. By modeling spatial configurations, visual patterns are more discriminative than local features, and preserve flexibility to tackle with object scaling, rotation, and deformation. We transfer the pattern discovery problem into finding frequent subgraphs from a graph, and exploit a graph mining algorithm to solve this problem, regardless of whether visual patterns are regularly arranged in images. We apply visual patterns to architecture image classification and product image retrieval, based on the ideas that visual pattern can describe elements conveying architecture styles and emblematic motifs of brands. Experimental results show that our pattern discovery approach has promising performance and is superior to the conventional bag-of-words approach.

2. Dataset

All the following images were downloaded from the internet, and are just used for academic research purpose.

Dataset for architecture image classification: [Download] (622 MB)

Dataset for product image search: [Download] (114MB)

3. Citation

W.-T. Chu and M.-H. Tsai, ¡§Visual Pattern Discovery for Architecture Image Classification and Product Image Search,¡¨ACM International Conference on Multimedia Retrieval, 2012.

Any problem please contact .