Basic Understanding __Convolutional Neural Networks Layers

What is a Convolutional Neural Network:

In deep learning Convolutional Neural Network (CNN or ConvNet) is a class of deep neural networks most commonly applied to analyse visual imagery.

In simple deep learning recognize the object in an image using a CNN algorithm.

Layers in Convolutional Neural Network:

Input Layer:

The pixels of the image will be fed as a matrix to the input layer

Hidden Layer:

The hidden layer contains the convolutional layer, ReLU layer and Pooling.

•     Convolutional layer - Performs convolution operation to detect the patterns in the image                                                  using the Matrix filter
•     ReLU layer              - Applied to convolution layer to get a rectified feature map of image
•     Pooling  layer          -  Perform features extraction from the image ( ex: like beak, feather,                                                  tail,edges, corners etc., in case of a bird image)

How Convolution Network recognizes images?

The images are considered as pixels and represented in the form of matrix by CNN

How Convolution operation happens in CNN?

Every image is considered as a matrix of pixel values.

Consider:

The original image(x) 6X6 is nXn

Convolution(w)  3X3 is fXf

Convolved feature(z) 4X4 is (n-f+1) = (6-3+1) = 4

Padding:

• Padding is used to preserve the original dimension of the input when ever any information given in the border
• Zeros are added outside of the input
• Number of zero layers depends on the size of the filter

Stride:

The stride can be used when scanning a large pixel image to skip the pixels.

The number of pixels shifts over the input matrix. When the stride is 2 then it will move the filters to 2 pixels at a time.

Pooling layer:

It is a function progressively reduce the spatial size of the representation to reduce the amount of parameter and computation network.

Types of Padding

• Max Padding
• Min Padding
• Average Padding

Fully Connected layer -  Have full connection to all activation in the previous layers

Here we go with an Example to classify an image in Keras

Lets take several images of cats and dogs using CNN which needs to see the image and understand by the various features as identify whether its a cat or dog.

First would need to resize the images to get them all in the same shape. This is the paramount act need to do while handling images as it would be impossible to capture all images of the same size,

For simplicity the author used a single layer and a single pooling layer, 

Click for the dataset here.

#import various packages

import os
import numpy as np
import pandas as pd
import scipy
import sklearn
import keras
from keras.models import Sequential
import cv2
from skimage import io
%matplotlib inline

#Defining the File Path

cat=os.listdir("/mnt/hdd/datasets/dogs_cats/train/cat")
dog=os.listdir("/mnt/hdd/datasets/dogs_cats/train/dog")
filepath="/mnt/hdd/datasets/dogs_cats/train/cat/"
filepath2="/mnt/hdd/datasets/dogs_cats/train/dog/"

#Loading the Images

images=[]
label = []
for i in cat:
    image = scipy.misc.imread(filepath+i)
    images.append(image)
    label.append(0) #for cat images

for i in dog:
    image = scipy.misc.imread(filepath2+i)
    images.append(image)
    label.append(1) #for dog images

#resizing all the images

for i in range(0,23000):
    images[i]=cv2.resize(images[i],(300,300))

#converting images to arrays

images=np.array(images)
label=np.array(label)

# Defining the hyperparameters

filters=10
filtersize=(5,5)

epochs =5
batchsize=128

input_shape=(300,300,3)

#Converting the target variable to the required size

from keras.utils.np_utils import to_categorical
label = to_categorical(label)

#Defining the model

model = Sequential()

model.add(keras.layers.InputLayer(input_shape=input_shape))

model.add(keras.layers.convolutional.Conv2D(filters, filtersize, strides=(1, 1), padding='valid', data_format="channels_last", activation='relu'))
model.add(keras.layers.MaxPooling2D(pool_size=(2, 2)))
model.add(keras.layers.Flatten())

model.add(keras.layers.Dense(units=2, input_dim=50,activation='softmax'))

model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
model.fit(images, label, epochs=epochs, batch_size=batchsize,validation_split=0.3)

model.summary()

References:

Simplilearn
Sridhar Swaminathan
Analytics Vidhya