Convolutional Neural Networks

Overview

Tensor Flow can be used to create convolutional neural networks.

A convolutional neural network is just a standard neural network, but includes a few extra layer types.

• Convolution Layers - implements a convolution on image data
• Pooling Layers - downsamples the image data
• Flatten Layer - flattens multi-dimensional data into 1-d array data.

Convolution Layers

model.add(tf.layers.conv2d({
    inputShape: [IMAGE_WIDTH, IMAGE_HEIGHT, IMAGE_CHANNELS],
    kernelSize: 5,
    filters: 8,
    strides: 1,
    activation: 'relu',
    kernelInitializer: 'varianceScaling'
  }));
				

Max Pooling Layers

 model.add(tf.layers.maxPooling2d({poolSize: [2, 2], strides: [2, 2]}));
				

Flatten Layers

model.add(tf.layers.flatten());