Image classification using convolution neural network with Keras − Blog by dchandra

This topic explains the image classification using CNN with Keras.

import numpy as np
import tensorflow as tf
import pandas as pd
import matplotlib.pyplot as plt
import random
import keras

from keras import models
from keras.layers import Dense, Conv2D, MaxPooling2D, Dropout, Flatten
from keras.utils import to_categorical
from keras.datasets import mnist

img_size = 28
n_class = 10
batch_size = 32
epochs = 10
input_shape = (img_size,img_size,1)

# Load data
(X_train, y_train), (X_test, y_test) = mnist.load_data()

X_train = X_train[:10000,:,:]
X_test = X_test[:1000,:,:]
y_train = y_train[:10000]
y_test  = y_test[:1000]

# Reshape data
X_train = X_train.reshape((X_train.shape[0], img_size,img_size, 1))
X_train = X_train.astype('float32') / 255
X_test = X_test.reshape((X_test.shape[0], img_size,img_size, 1))
X_test = X_test.astype('float32') / 255

# Categorically encode labels
y_train = to_categorical(y_train, n_class)
y_test = to_categorical(y_test, n_class)

# Build neural network
model = models.Sequential()

# First Convolution Layer

model.add(Conv2D(filters=16,kernel_size=(5,5),strides=(1, 1), padding='same',
                 activation='relu',kernel_initializer='random_uniform',bias_initializer='zeros',input_shape=input_shape))
model.add(MaxPooling2D(pool_size=(2, 2),strides=(2,2),padding='valid'))
model.add(Dropout(0.2))

# Second Convolution Layer

model.add(Conv2D(filters=32,kernel_size=(5,5),strides=(1, 1), padding='same',
                 activation='relu',kernel_initializer='random_uniform',bias_initializer='zeros'))
model.add(MaxPooling2D(pool_size=(2, 2),strides=(2,2),padding='valid'))
model.add(Dropout(0.2))

# Third Convolution Layer

model.add(Conv2D(filters=64,kernel_size=(5,5),strides=(1, 1), padding='same',
                 activation='relu',kernel_initializer='random_uniform',bias_initializer='zeros'))
model.add(MaxPooling2D(pool_size=(2, 2),strides=(2,2),padding='valid'))
model.add(Dropout(0.2))

# Fouth Convolution Layer

model.add(Conv2D(filters=128,kernel_size=(5,5),strides=(1, 1), padding='same',
                 activation='relu',kernel_initializer='random_uniform',bias_initializer='zeros'))
model.add(MaxPooling2D(pool_size=(2, 2),strides=(2,2),padding='valid'))
model.add(Dropout(0.2))

# fully connected layer

model.add(Flatten())
model.add(Dense(128, activation='relu'))
model.add(Dropout(0.3))

# output layer

model.add(Dense(n_class, activation='softmax'))

# Compile model
opt= keras.optimizers.Adagrad(lr=0.001, epsilon=None, decay=0.0)
model.compile(optimizer=opt,
              loss='categorical_crossentropy',
              metrics=['accuracy'])

model.fit(X_train, y_train,
          batch_size=batch_size,
          epochs=epochs,verbose = 1,
          validation_data=(X_test, y_test))

Train on 10000 samples, validate on 1000 samples
Epoch 1/10
10000/10000 [==============================] - 22s 2ms/step - loss: 1.6401 - acc: 0.4449 - val_loss: 0.8522 - val_acc: 0.7430
Epoch 2/10
10000/10000 [==============================] - 21s 2ms/step - loss: 0.9093 - acc: 0.7138 - val_loss: 0.6227 - val_acc: 0.8170
Epoch 3/10
10000/10000 [==============================] - 21s 2ms/step - loss: 0.7399 - acc: 0.7683 - val_loss: 0.5140 - val_acc: 0.8530
Epoch 4/10
10000/10000 [==============================] - 21s 2ms/step - loss: 0.6561 - acc: 0.7965 - val_loss: 0.4454 - val_acc: 0.8740
Epoch 5/10
10000/10000 [==============================] - 21s 2ms/step - loss: 0.5913 - acc: 0.8184 - val_loss: 0.4005 - val_acc: 0.8860
Epoch 6/10
10000/10000 [==============================] - 21s 2ms/step - loss: 0.5659 - acc: 0.8280 - val_loss: 0.3751 - val_acc: 0.8970
Epoch 7/10
10000/10000 [==============================] - 21s 2ms/step - loss: 0.5123 - acc: 0.8454 - val_loss: 0.3458 - val_acc: 0.9010
Epoch 8/10
10000/10000 [==============================] - 21s 2ms/step - loss: 0.4843 - acc: 0.8517 - val_loss: 0.3303 - val_acc: 0.9010
Epoch 9/10
10000/10000 [==============================] - 21s 2ms/step - loss: 0.4587 - acc: 0.8552 - val_loss: 0.3123 - val_acc: 0.9100
Epoch 10/10
10000/10000 [==============================] - 21s 2ms/step - loss: 0.4418 - acc: 0.8665 - val_loss: 0.2954 - val_acc: 0.9160

score = model.evaluate(X_test, y_test, verbose=0)
print('Test loss:', score[0])
print('Test accuracy:', score[1])

# Summary of neural network
model.summary()

Test loss: 0.29542376232147216
Test accuracy: 0.916
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_1 (Conv2D)            (None, 28, 28, 16)        416       
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 14, 14, 16)        0         
_________________________________________________________________
dropout_1 (Dropout)          (None, 14, 14, 16)        0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 14, 14, 32)        12832     
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 7, 7, 32)          0         
_________________________________________________________________
dropout_2 (Dropout)          (None, 7, 7, 32)          0         
_________________________________________________________________
conv2d_3 (Conv2D)            (None, 7, 7, 64)          51264     
_________________________________________________________________
max_pooling2d_3 (MaxPooling2 (None, 3, 3, 64)          0         
_________________________________________________________________
dropout_3 (Dropout)          (None, 3, 3, 64)          0         
_________________________________________________________________
conv2d_4 (Conv2D)            (None, 3, 3, 128)         204928    
_________________________________________________________________
max_pooling2d_4 (MaxPooling2 (None, 1, 1, 128)         0         
_________________________________________________________________
dropout_4 (Dropout)          (None, 1, 1, 128)         0         
_________________________________________________________________
flatten_1 (Flatten)          (None, 128)               0         
_________________________________________________________________
dense_1 (Dense)              (None, 128)               16512     
_________________________________________________________________
dropout_5 (Dropout)          (None, 128)               0         
_________________________________________________________________
dense_2 (Dense)              (None, 10)                1290      
=================================================================
Total params: 287,242
Trainable params: 287,242
Non-trainable params: 0
_________________________________________________________________

X_test = X_test.reshape((X_test.shape[0], img_size,img_size, 1))

indices = random.sample(range(X_test.shape[0]),k=10)
X_test_samples = X_test[indices]
y_test_samples = y_test[indices]

for i in range(10):
  predict = model.predict_classes(X_test_samples[i].reshape((1,img_size,img_size, 1)))
  actual = np.argmax(y_test_samples[i])
  plt.imshow(X_test_samples[i].reshape(28,28),cmap="binary")
  print("Actual Class :{}, Predicted Class: {}".format(actual,predict[0]))
  plt.show()