Multi-feature fusion based spatial pyramid deep neural networks image classification

Qingyong Xu^{1, 2}, Shunliang Jiang², Wei Huang², Longzhen Duan², Shaoping Xu²

COMPUTER MODELLING & NEW TECHNOLOGIES 2015 19(2C) 16-21

¹School of Economic and Management, Nanchang University, Nanchang 330031, China
²School of Information Engineering, Nanchang University, Nanchang 330031, China

The scalable and efficient multi-class classification algorithm is now a well-known hard problem. Traditional methods of computer vision and machine learning cannot match human performance on images classification tasks. This paper proposes a novel semi-supervised classifier called Spatial Pyramid Deep Neural Networks (SPDNN). SPDNN utilizes a new deep architecture to integrate the ability of neural networks and spatial pyramid model because deep neural networks do not considerable the spatial information. Feature fusion has been more and more important for image and video retrieval, indexing and annotation because of the lack of single feature. We use multiple feature fusion over any single feature instead of pixels of images. The features include color feature, shape feature and texture feature. The performance of experiment shows that the algorithm improved the state-of-the-art image classification.