Title: Asymmetric convolution with densely connected networks

Authors: Liejun Wang; Huanglu Wen; Jiwei Qin; Shuli Cheng

Addresses: School of Information Science and Engineering, Xinjiang University, Urumqi 830046, China; School of Software, Xinjiang University, Urumqi 830046, China ' School of Information Science and Engineering, Xinjiang University, Urumqi 830046, China ' School of Education, Shaanxi Normal University, Xi'an 710062, China ' School of Information Science and Engineering, Xinjiang University, Urumqi 830046, China

Abstract: Convolutional neural networks are vital to some computer vision tasks, and the densely connected network is a creative architecture among them. In densely connected network, most convolution layer tends to have a much larger number of input channels than output channels, making itself to a funnel shape. We replace the 3 × 3 convolution in the densely connected network with two continuous asymmetric convolutions to make the DenseNet family more diverse. We also proposed a model in which two continuous asymmetric convolutions each outputs half of the output channels and concatenate them as the final output of these layers. Compared with the original densely connected network, our models achieve similar performance on CIFAR-10/100 dataset with fewer parameters and less computational cost.

Keywords: densely connected network; DenseNet; asymmetric convolution; concatenation.

DOI: 10.1504/IJCSM.2020.111704

International Journal of Computing Science and Mathematics, 2020 Vol.12 No.3, pp.274 - 284

Received: 30 Jul 2018
Accepted: 10 Sep 2018
Published online: 11 Dec 2020 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article