![\"Keras](\"https:\/\/data-flair.training\/blogs\/wp-content\/uploads\/sites\/2\/2020\/07\/Keras-Convolution-Neural-Network.jpg\")
<\/a><\/p>\nConvolution Neural Network Architecture<\/h2>\n
The convolution neural network algorithm is the result of continuous advancements in computer vision with deep learning.
\nCNN is a Deep learning algorithm that is able to assign importance to various objects in the image and able to differentiate them.
\nCNN has the ability to learn the characteristics and perform classification.
\nAn input image has many spatial and temporal dependencies, CNN captures these characteristics using relevant filters\/kernels.<\/p>\n
A Kernel or filter is an element in CNN that performs convolution around the image in the first part. The kernel moves to the right and shifts according to the stride value. Every time during convolution a matrix multiplication operation is performed.<\/p>\n
After convolution, we obtain another image with a different height, width, and depth. We obtain more channels than just RGB but less width and height.<\/p>\n
We slide each filter though out the image step by step, this step in the forward pass is called stride.<\/p>\n
Layers in CNN<\/h2>\n1. Keras Convolution layer<\/h3>\n
It is the first layer to extract features from the input image. Here we define the kernel as the layer parameter. We perform matrix multiplication operations on the input image using the kernel.<\/p>\n
Example:<\/p>\n
Suppose a 3*3 image pixel and a 2*2 filter as shown:<\/p>\n
pixel : [[1,0,1],
\n[0,1,0],
\n[1,0,1]]
\nfilter : [[1,0],
\n[0,1]]<\/p>\n
The restaurant matrix after convolution of filter would be:
\n[[2,0],
\n[0,2]]<\/p>\n
After convolution, we perform pooling to reduce the number of parameters and computations. Example: Image pixels: The resultant matrix after max-pooling would be: It is used to prevent the network from overfitting. In this layer, some fraction of units in the network is dropped in training such that the model is trained on all the units.<\/p>\n A series of convolution and pooling layers are used for feature extraction. After that, we construct densely connected layers to perform classification based on these features.<\/p>\n It is used to convert the data into 1D arrays to create a single feature vector. After flattening we forward the data to a fully connected layer for final classification.<\/p>\n It is a fully connected layer. Each node in this layer is connected to the previous layer i.e densely connected. This layer is used at the final stage of CNN to perform classification.<\/p>\n CIFAR 10 dataset consists of 10 image classes. The available image classes are :<\/p>\n This is one of the most popular datasets that allow researchers to practice different algorithms for object recognition. Let’s build our Convolution model to recognize CIFAR-10 classes.<\/p>\n 1. Load the dataset from keras datasets module.<\/strong><\/p>\n 2. To visualize the dataset<\/strong><\/p>\n 3. The dataset looks like<\/strong><\/p>\n 4. Normalizing inputs<\/strong><\/p>\n 5. One hot encoding<\/strong><\/p>\n 6. Build the model<\/strong><\/p>\n 7. Model Compiling<\/strong><\/p>\n 8. Analyzing Model Summary<\/strong><\/p>\n 9. Train the model and check its accuracy on test data<\/strong><\/p>\n 10. Evaluate the model<\/strong><\/p>\n The Fashion MNIST dataset consists of a training set of 60000 images and a testing set of 10000 images. There are 10 image classes in this dataset and each class has a mapping corresponding to the following labels:<\/p>\n Let’s build our CNN model on this dataset.<\/p>\n 1. Import required modules<\/strong><\/p>\n 2. Load the dataset<\/strong><\/p>\n 3. Reshaping and one hot encoding<\/strong><\/p>\n 4. Visualize the dataset using matplotlib<\/strong><\/p>\n<\/a><\/p>\n2. Keras Pooling Layer<\/h3>\n
\nThere are different types of pooling operations, the most common ones are max pooling and average pooling.<\/p>\n
\nTake a sample case of max pooling with 2*2 filter and stride 2.<\/p>\n
\n[[1,2,3,4],
\n[5,6,7,8],
\n[3,4,5,6],
\n[6,7,8,9]]<\/p>\n
\n[[6,8],
\n[7,9]]<\/p>\n<\/a>
\n3. Keras Dropout Layer<\/h3>\n<\/a><\/p>\n4. Keras Flatten Layer<\/h3>\n
<\/a><\/p>\n5. Keras Dense Layer<\/h3>\n
<\/a><\/p>\nImplementing CNN on CIFAR 10 Dataset<\/h2>\n
\n
\nConvolution Neural Networks have shown the best results in solving the CIFAR-10 problem.<\/p>\nfrom keras.datasets import cifar10\r\nimport matplotlib.pyplot as plt\r\n\r\n(train_X,train_Y),(test_X,test_Y)=cifar10.load_data()<\/pre>\n
n=6\r\nplt.figure(figsize=(20,10))\r\nfor i in range(n):\r\nplt.subplot(330+1+i)\r\nplt.imshow(train_X[i])\r\nplt.show()<\/pre>\n
<\/a>
\nNow import required modules,<\/p>\nfrom keras.models import Sequential\r\nfrom keras.layers import Dense\r\nfrom keras.layers import Dropout\r\nfrom keras.layers import Flatten\r\nfrom keras.constraints import maxnorm\r\nfrom keras.optimizers import SGD\r\nfrom keras.layers.convolutional import Conv2D\r\nfrom keras.layers.convolutional import MaxPooling2D\r\nfrom keras.utils import np_utils<\/pre>\n
train_x=train_X.astype('float32')\r\ntest_X=test_X.astype('float32')\r\n\r\ntrain_X=train_X\/255.0\r\ntest_X=test_X\/255.0<\/pre>\ntrain_Y=np_utils.to_categorical(train_Y)\r\ntest_Y=np_utils.to_categorical(test_Y)\r\n\r\nnum_classes=test_Y.shape[1]<\/pre>\n
model=Sequential()\r\nmodel.add(Conv2D(32,(3,3),input_shape=(32,32,3),padding='same',activation='relu',kernel_constraint=maxnorm(3)))\r\nmodel.add(Dropout(0.2))\r\nmodel.add(Conv2D(32,(3,3),activation='relu',padding='same',kernel_constraint=maxnorm(3)))\r\nmodel.add(MaxPooling2D(pool_size=(2,2)))\r\nmodel.add(Flatten())\r\nmodel.add(Dense(512,activation='relu',kernel_constraint=maxnorm(3)))\r\nmodel.add(Dropout(0.5))\r\nmodel.add(Dense(num_classes, activation='softmax'))<\/pre>\n
sgd=SGD(lr=0.01,momentum=0.9, decay=(0.01\/25),nesterov=False)\r\nmodel.compile(loss='categorical_crossentropy',optimizer=sgd,metrics=['accuracy'])<\/pre>\n
model.summary()<\/pre>\n
<\/a><\/p>\nmodel.fit(train_X,train_Y,validation_data=(test_X,test_Y),epochs=10,batch_size=32)<\/pre>\n
<\/a><\/p>\n_,acc=model.evaluate(test_X,test_Y)\r\nprint(acc*100)<\/pre>\n
Implementing CNN on Fashion MNIST Dataset<\/h2>\n
\n
from numpy import mean\r\nfrom numpy import std\r\nfrom matplotlib import pyplot\r\nfrom sklearn.model_selection import KFold\r\nfrom keras.datasets import fashion_mnist\r\nfrom keras.utils import to_categorical\r\nfrom keras.models import Sequential\r\nfrom keras.layers import Conv2D\r\nfrom keras.layers import MaxPooling2D\r\nfrom keras.layers import Dense\r\nfrom keras.layers import Flatten\r\nfrom keras.optimizers import SGD<\/pre>\n
(train_X,train_Y),(test_X,test_Y)=fashion_mnist.load_data()<\/pre>\n
train_X=train_X.reshape((train_X.shape[0],28,28,1))\r\ntest_X=test_X.reshape((test_X.shape[0],28,28,1))\r\n\r\ntrain_Y=to_categorical(train_Y)\r\ntest_Y=to_categorical(test_Y)<\/pre>\n
n=5\r\n\r\npyplot.figure(figsize=(20,2))\r\n\r\nfor i in range(n):\r\nax=pyplot.subplot(1,n,i+1)\r\npyplot.imshow(train_X[i].reshape(28,28))\r\npyplot.gray()\r\nax.get_xaxis().set_visible(False)\r\nax.get_yaxis().set_visible(False)\r\npyplot.show()<\/pre>\n