Abstract:In this paper, we propose a deep neural network architecture for object recognition based on recurrent neural networks. The proposed network, called ReNet, replaces the ubiquitous convolution+pooling layer of the deep convolutional neural network with four recurrent neural networks that sweep horizontally and vertically in both directions across the image. We evaluate the proposed ReNet on three widely-used benchmark datasets; MNIST, CIFAR-10 and SVHN. The result suggests that ReNet is a viable alternative to the deep convolutional neural network, and that further investigation is needed.
Submission history
From: Francesco Visin [view email]
[v1]
Sun, 3 May 2015 04:58:53 UTC (797 KB)
[v2]
Fri, 3 Jul 2015 11:31:53 UTC (798 KB)
[v3]
Thu, 23 Jul 2015 17:11:04 UTC (798 KB)