Quaddirectional 2D-Recurrent Neural Networks For Image Labeling

Author

Bing Shuai ; Zhen Zuo ; Gang Wang

Author_Institution

Sch. of Electr. & Electron. Eng., Nanyang Technol. Univ., Singapore, Singapore

Volume

22

Issue

11

fYear

2015

fDate

Nov. 2015

Firstpage

1990

Lastpage

1994

Abstract

We adopt Convolutional Neural Networks (CNN) to learn discriminative features for local patch classification. We further introduce quaddirectional 2D Recurrent Neural Networks to model the long range dependencies among pixels. Our quaddirectional 2D-RNN is able to embed the global image context into the compact local representation, which significantly enhance their discriminative power. Our experiments demonstrate that the integration of CNN and quaddirectional 2D-RNN achieves very promising results which are comparable to state-of-the-art on real-world image labeling benchmarks.

Keywords

feature extraction; image classification; learning (artificial intelligence); recurrent neural nets; CNN; compact local representation; convolutional neural networks; discriminative feature learning; discriminative power enhancement; global image context; image labeling; image pixels; local patch classification; long-range dependency model; quaddirectional 2D-RNN; quaddirectional 2D-recurrent neural networks; real-world image labeling benchmarks; Context; Convolution; Feature extraction; Hidden Markov models; Labeling; Recurrent neural networks; Convolutional neural networks; image labeling; quaddirectional 2D-RNN; recurrent neural networks;

fLanguage

English

Journal_Title

Signal Processing Letters, IEEE

Publisher

ieee

ISSN

1070-9908

Type

jour

DOI

10.1109/LSP.2015.2441781

Filename

7118156