Applying CNN on MNIST Dataset

Importing Necessary Libraries 

 

Firstly, we will load some basic libraries:-

 

(i) Numpy - for linear algebra. 

 

(ii) Pandas - for data analysis. 

 

(iii) Matplotlib - for data visualization.

 

(iv) Tensorflow - for neural networks.

 

import numpy as np import pandas as pd  from numpy import unique, argmax from tensorflow.keras.datasets.mnist import load_data  from tensorflow.keras import Sequential from tensorflow.keras.layers import Conv2D from tensorflow.keras.layers import Dense  from tensorflow.keras.layers import Flatten  from tensorflow.keras.layers import Dropout  from tensorflow.keras.utils import plot_model import matplotlib.pyplot as plt from tensorflow.keras.datasets import mnist

 

Loading Data 

 

We will load the MNIST Dataset using the Keras library and split it into training and testing sets.

#loading the MNIST Dataset  (train_x, train_y), (test_x, test_y) = mnist.load_data()

 

Now, let us print the dimensions of the dataset. 

#printing the shapes  print(train_x.shape, train_y.shape) print(test_x.shape , test_y.shape)

 

Output

We notice 60,000 images in the train set and 10,000 images in the test set.

Visualization and Preprocessing

 

Now, we will reshape the datasets since the images are greyscaled. 

reshaping train and test sets  train_x = train_x.reshape((train_x.shape[0], train_x.shape[1], train_x.shape[2], 1)) test_x = test_x .reshape((test_x.shape[0], test_x.shape[1], test_x.shape[2], 1))

 

Now, let us print the new dimensions. 

#printing the shapes  print(train_x.shape, train_y.shape) print(test_x.shape , test_y.shape)

 

Output

 

 

Now, we will perform normalization for scaling the image pixels. This will help improve the model's performance because the value will range from 0 to 1 instead of 0 to 255.  

#normalizing the pixel values of images  train_x = train_x.astype('float32')/255.0 test_x = test_x.astype('float32')/255.0

 

Now, let us plot the first 20 images for better visualization. 

#plotting images of dataset  fig = plt.figure(figsize = (10,3)) for i in range(20):     ax= fig.add_subplot(2, 10, i+1, xticks=[], yticks=[])     ax.imshow(np.squeeze(train_x[i]), cmap='gray')     ax.set_title(train_y[i])

 

Output

 

 

Let us try to print the shape of a single image. 

shape = train_x.shape[1:] shape

 

Output

 

 

Here, 1 indicates the channel of the image.

 

Training 

 

Now, let us build the CNN model by adding significant layers. 

#CNN Model  model = Sequential() #adding convolutional layer  model.add(Conv2D(32, (3,3), activation='relu', input_shape= shape)) model.add(MaxPool2D((2,2))) model.add(Conv2D(48, (3,3), activation='relu')) model.add(MaxPool2D((2,2))) model.add(Dropout(0.5)) model.add(Flatten()) model.add(Dense(500, activation='relu')) model.add(Dense(10, activation='softmax'))

 

Let us print a summary of our model for better understanding. 

model.summary()

 

Output

 

 

Now let us compile the model and train it. 

#compiling model  model.compile(optimizer='adam', loss = 'sparse_categorical_crossentropy',metrics= ['accuracy'] ) x=model.fit(train_x, train_y, epochs=10, batch_size = 128, verbose= 2 , validation_split = 0.1)

 

Results 

 

Finally, let us evaluate our model. 

loss, accuracy= model.evaluate(test_x, test_y, verbose = 0) print(f'Accuracy: {accuracy*100}')

 

Output

 

 

FAQs 

1. What is a Deep Learning?
   Deep learning is closely related to machine learning, but it learns to perform classification tasks directly from images, text, or sound. 
    
2. What is the difference between ANN and CNN?
   ANN or Artificial Neural Network is a multi-layer fully-connected neural net that consists of many layers, including an input layer, multiple hidden layers, and an output layer. CNNs are neural nets that share their parameters. A covnet is a sequence of layers, and each layer transforms one volume to another through a differentiable function.

 

Key Takeaways 

 

Congratulations on making it this far. This blog discussed a fundamental overview of CNN and the MNIST Dataset !!

 

We learned about Data Loading and Data Visualisation, and finally, we applied CNN on the MNIST Dataset. 

Check out this article - Padding In Convolutional Neural Network

If you are preparing for the upcoming Campus Placements, don’t worry. Coding Ninjas has your back. Visit this link for a carefully crafted and designed course on-campus placements and interview preparation.