residual-network

Deep Residual Network

You can use this library to make your own ResNet.
It is very customizable and use TensorFlow.

Installation

The library requires Python 3.6.
Installation is very simple. You can use PIP.

pip3 install resnet

If you want to install ResNet from source..

python3.6 setup.py install

Deep Residual Neural Network (ResNet) Example

Run ResNet

The Git codes contains CIFAR-10 image classification example.
All you need to do is very simple.

python3.6 main.py --mode=train

Create Your Own ResNet

You might want to customize or make your own ResNet.
The following code shows you how to make your own ResNet.

resnet = ResNet(batch=32)
with tf.variable_scope('input_scope'):
    h = resnet.init_block(filter=[7, 7], channel=[3, 32], max_pool=False)

with tf.variable_scope('residual01'):
    h = resnet.max_pool(h, kernel=[2, 2], stride=[2, 2])
    h = resnet.residual_block(h, filter=[3, 3], channel=[32, 32])
    h = resnet.residual_block(h, filter=[3, 3], channel=[32, 32])
    h = resnet.residual_block(h, filter=[3, 3], channel=[32, 32])
    h = resnet.residual_block(h, filter=[3, 3], channel=[32, 32])
    h = resnet.residual_block(h, filter=[3, 3], channel=[32, 32])
    h = resnet.residual_block(h, filter=[3, 3], channel=[32, 32])

with tf.variable_scope('residual02'):
    h = resnet.max_pool(h, kernel=[2, 2], stride=[2, 2])
    h = resnet.residual_block(h, filter=[3, 3], channel=[32, 64])
    h = resnet.residual_block(h, filter=[3, 3], channel=[64, 64])
    h = resnet.residual_block(h, filter=[3, 3], channel=[64, 64])
    h = resnet.residual_block(h, filter=[3, 3], channel=[64, 64])
    h = resnet.residual_block(h, filter=[3, 3], channel=[64, 64])
    h = resnet.residual_block(h, filter=[3, 3], channel=[64, 64])
    h = resnet.residual_block(h, filter=[3, 3], channel=[64, 64])
    h = resnet.residual_block(h, filter=[3, 3], channel=[64, 64])

with tf.variable_scope('residual03'):
    h = resnet.max_pool(h, kernel=[2, 2], stride=[2, 2])
    h = resnet.residual_block(h, filter=[3, 3], channel=[64, 128])
    h = resnet.residual_block(h, filter=[3, 3], channel=[128, 128])
    h = resnet.residual_block(h, filter=[3, 3], channel=[128, 128])
    h = resnet.residual_block(h, filter=[3, 3], channel=[128, 128])
    h = resnet.residual_block(h, filter=[3, 3], channel=[128, 128])
    h = resnet.residual_block(h, filter=[3, 3], channel=[128, 128])
    h = resnet.residual_block(h, filter=[3, 3], channel=[128, 128])

with tf.variable_scope('residual04'):
    h = resnet.max_pool(h, kernel=[2, 2], stride=[2, 2])
    h = resnet.residual_block(h, filter=[3, 3], channel=[128, 256])
    h = resnet.residual_block(h, filter=[3, 3], channel=[256, 256])
    h = resnet.residual_block(h, filter=[3, 3], channel=[256, 256])
    h = resnet.residual_block(h, filter=[3, 3], channel=[256, 256])
    h = resnet.residual_block(h, filter=[3, 3], channel=[256, 256])

with tf.variable_scope('residual05'):
    h = resnet.max_pool(h, kernel=[2, 2], stride=[2, 2])
    h = resnet.residual_block(h, filter=[3, 3], channel=[256, 512])
    h = resnet.residual_block(h, filter=[3, 3], channel=[512, 512])
    h = resnet.residual_block(h, filter=[3, 3], channel=[512, 512])
    h = resnet.residual_block(h, filter=[3, 3], channel=[512, 512])
    h = resnet.residual_block(h, filter=[3, 3], channel=[512, 512])
    h = resnet.residual_block(h, filter=[3, 3], channel=[512, 512])

with tf.variable_scope('fc'):
    h = resnet.avg_pool(h, kernel=[2, 2], stride=[2, 2])
    h = resnet.fc(h)
h # <- Your Network Created