Skip to content

Keras implementation of influential CNN architectures.

Notifications You must be signed in to change notification settings

s-themis/cnn-architectures

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

12 Commits
 
 
 
 
 
 
 
 

Repository files navigation

cnn-architectures

Implementation of influential convolutional neural network architectures using the Keras functional API.

Architectures

AlexNet

Paper | Implementation
Year published: 2012
Depth: 8
Parameters: 62.4M
Reported training time: 5-6 days on two GTX 580 3GB GPUs
Key architectural points:

  • Kernels of decreasing size 11x11, 5x5, 3x3
  • ReLU activations
  • Local response normalization
  • Overlapping max pooling for downsampling
  • Dropout on fully-connected layers

VGGNet (Configuration D - VGG16)

Paper | Implementation
Year published: 2014
Depth: 16
Parameters: 138.4M
Reported training time: 2-3 weeks on four Titan Black GPUs
Key architectural points:

  • Homogeneous convolutional blocks with decreasing feature map size and increasing number of filters
  • Homogeneous kernels of size 3x3
  • ReLU activations
  • No local response normalization
  • Non-overlapping max pooling for downsampling on each block
  • Dropout on fully-connected layers

ResNet (ResNet50 Variant)

Paper | Implementation
Year published: 2015
Depth: 50
Parameters: 25.6M
Reported training time: -
Key architectural points:

  • Initial kernel size 7x7 followed by overlapping max-pooling for downsampling
  • Homogenous stacks of residual blocks with decreasing feature map size and increasing number of filters
  • Kernels of size 1x1, 3x3, 1x1 on each residual block to bottleneck number of filters on 3x3 layers
  • ReLU activations
  • Batch normalization before activations
  • Downsampling with overlapping convolutions
  • Global average pooling before single fully-connected layer
  • No dropout

Releases

No releases published

Packages

No packages published

Languages