Residual Networks of Residual Networks in Keras

This is an implementation of the paper "Residual Networks of Residual Networks: Multilevel Residual Networks"

Explanation

Ordinarily, Residual networks have hundreds or even thousands of layers to accurately classify images in major image recognition tasks, but building a network by simply stacking residual blocks inevitably limits its optimization ability.

This paper attempts to improve the optimization ability of Residual Networks by adding level-wise shortcut connections upon original residual networks, to promote the learning capability of residual networks.

This can be shown by the figure from the paper :

There are two different architectures available, since RoR can be extended to Wide Residual Networks or Pre-ResNets as well.

The two images below from the paper describe the architecture of RoR on ResNets and RoR on Wide Residial Networks:

The classification accuracy of these networks on CIFAR 10 (from the paper) are :

Usage

The paper uses several models such as RoR-3-110 (ResNet-101) and and RoR-3-WRN-40-2 (Wide Residual Network-40-2) models but due to GPU memory limitations, only the weights for the RoR-3-WRN-40-2 have been provided in the Releases tab

Please download the weights and place them in the weights folder.

To create RoR ResNet models, use the ror.py script :

import ror

input_dim = (3, 32, 32) if K.image_dim_ordering() == 'th' else (32, 32, 3)
model = ror.create_residual_of_residual(input_dim, nb_classes=100, N=2, dropout=0.0) # creates RoR-3-110 (ResNet)

To create RoR Wide Residual Network models, use the ror_wrn.py script :

import ror_wrn as ror

input_dim = (3, 32, 32) if K.image_dim_ordering() == 'th' else (32, 32, 3)
model = ror.create_pre_residual_of_residual(input_dim, nb_classes=100, N=6, k=2, dropout=0.0) # creates RoR-3-WRN-40-2 (WRN)

Performance

The RoR-WRN-40-2 model described in the paper requires 500 epochs to acheive a classification accuracy of 94.99 % (5.01 % error).

The Theano weights provided for this model are trained for 100 epochs using Adam with a learning rate of 1e-3, which achieves a classification accuracy of 94.48% (5.52 % error)

Requirements

• Keras
• Theano (weights provided) / Tensorflow (weights not yet converted)
• scipy
• h5py
• sklearn (for metrics)