Image recognition using convolutional neural network

Face recognition using CNN

import os
import cv2
import numpy as np
import pandas as pd
import tensorflow as tf
import sklearn
from scipy import ndimage
from PIL import Image
from sklearn import preprocessing
import matplotlib.pyplot as plt
%matplotlib inline
from sklearn.utils import shuffle
from keras.models import Sequential
from keras.layers import Conv2D, Flatten, Dropout, Dense, MaxPooling2D, Cropping2D
from keras.optimizers import Adam
from keras.callbacks import EarlyStopping

data = 'yalefaces/'
faces_data = []
class_label = []
for filename in os.listdir(data):
    if filename.endswith(".txt") or filename.endswith(".DS_Store") :
        continue
    else:
        #extractFrames(filename, 'data_faces')
        img_array = np.array(Image.open(data+filename)).astype(float)
        faces_data.append(img_array)
        label = filename[7:9]
        class_label.append(label)
faces_data = np.array(faces_data)
class_label = np.array(class_label)

nb_class = len(np.unique(class_label))

lb = preprocessing.LabelBinarizer()
class_label = lb.fit_transform(class_label)

print(class_label)

[[0 0 0 ..., 0 0 0]
 [1 0 0 ..., 0 0 0]
 [0 0 0 ..., 0 0 0]
 ..., 
 [0 0 0 ..., 0 0 0]
 [0 0 0 ..., 1 0 0]
 [0 0 0 ..., 0 0 0]]

def normalize(img):
    return 1/255.*(img)

def crop(img):
    return img[:, 50:300]

X = []
for img in faces_data:
    #img = crop(img)
    img = normalize(img)
    img = cv2.resize(img, (160,121))
    X.append(img)
X = np.array(X)
print(X.shape)

(166, 121, 160)

from sklearn.utils import shuffle
X, y = shuffle(X, class_label, random_state=0)

plt.figure()
plt.imshow(X[10], cmap='gray')

<matplotlib.image.AxesImage at 0x119c7ce80>

alt text

from sklearn.model_selection import train_test_split

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.33)

# For input to Keras model: (batch size, height, width, channels)
X_train = np.reshape(X_train, (X_train.shape[0], X_train.shape[1], X_train.shape[2], 1))
X_test = np.reshape(X_test, (X_test.shape[0], X_test.shape[1], X_test.shape[2], 1))
print(X_train.shape, y_train.shape)
print(X_test.shape, y_test.shape)

(111, 121, 160, 1) (111, 15)
(55, 121, 160, 1) (55, 15)

model = Sequential()

model.add(Conv2D(16, (3,3), strides=(1,1), activation='elu', input_shape=(X_train.shape[1],X_train.shape[2],1)))
model.add(MaxPooling2D((2,2), strides=(2,2)))
model.add(Conv2D(32, (3,3), activation='elu'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
model.add(Conv2D(32, (3,3), strides=(1,1), activation='elu'))
model.add(MaxPooling2D((2,2), strides=(2,2)))
model.add(Flatten())
model.add(Dropout(0.5))
model.add(Dense(512, activation='elu'))
model.add(Dropout(0.3))
model.add(Dense(128, activation='elu'))
model.add(Dropout(0.3))
model.add(Dense(nb_class, activation='softmax' ))

model.summary()

model.compile(optimizer='Adam',
          loss='categorical_crossentropy',
          metrics=['accuracy'])
#model.fit(X_train, y_train, epochs=8, batch_size=32)

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_4 (Conv2D)            (None, 119, 158, 16)      160       
_________________________________________________________________
max_pooling2d_4 (MaxPooling2 (None, 59, 79, 16)        0         
_________________________________________________________________
conv2d_5 (Conv2D)            (None, 57, 77, 32)        4640      
_________________________________________________________________
max_pooling2d_5 (MaxPooling2 (None, 28, 38, 32)        0         
_________________________________________________________________
conv2d_6 (Conv2D)            (None, 26, 36, 32)        9248      
_________________________________________________________________
max_pooling2d_6 (MaxPooling2 (None, 13, 18, 32)        0         
_________________________________________________________________
flatten_2 (Flatten)          (None, 7488)              0         
_________________________________________________________________
dropout_4 (Dropout)          (None, 7488)              0         
_________________________________________________________________
dense_4 (Dense)              (None, 512)               3834368   
_________________________________________________________________
dropout_5 (Dropout)          (None, 512)               0         
_________________________________________________________________
dense_5 (Dense)              (None, 128)               65664     
_________________________________________________________________
dropout_6 (Dropout)          (None, 128)               0         
_________________________________________________________________
dense_6 (Dense)              (None, 15)                1935      
=================================================================
Total params: 3,916,015
Trainable params: 3,916,015
Non-trainable params: 0
_________________________________________________________________

Question 2 (f)

Data Augmentation

# Data Augmentation
from keras.preprocessing.image import ImageDataGenerator
EPOCHS = 30
BATCH_SIZE = 32
train_datagen = ImageDataGenerator(shear_range=0.2,
                                   zoom_range=0.2,
                                   horizontal_flip=True)

test_datagen = ImageDataGenerator()

train_generator = train_datagen.flow(X_train, y_train, batch_size=BATCH_SIZE)

validation_generator = test_datagen.flow(X_test, y_test, batch_size=BATCH_SIZE)

#earlyStopping = EarlyStopping(monitor='val_loss', min_delta=0, patience=3, verbose=1, mode='auto')

history = model.fit_generator(train_generator,
    steps_per_epoch=len(X_train) // BATCH_SIZE,
    epochs=EPOCHS,
    validation_data = validation_generator,
    validation_steps = len(X_test) // BATCH_SIZE,
    verbose=1)
    #callbacks=[earlyStopping])

model.save('model.h5')
print("Model saved!")

Epoch 1/30
4/3 [========================================] - 3s 861ms/step - loss: 3.4328 - acc: 0.0575 - val_loss: 2.9319 - val_acc: 0.0182
Epoch 2/30
4/3 [========================================] - 3s 686ms/step - loss: 2.9228 - acc: 0.1403 - val_loss: 2.5272 - val_acc: 0.1818
Epoch 3/30
4/3 [========================================] - 3s 749ms/step - loss: 2.4885 - acc: 0.1838 - val_loss: 2.4730 - val_acc: 0.1636
Epoch 4/30
4/3 [========================================] - 2s 611ms/step - loss: 2.2977 - acc: 0.3136 - val_loss: 2.2694 - val_acc: 0.3273
Epoch 5/30
4/3 [========================================] - 3s 742ms/step - loss: 1.9934 - acc: 0.4288 - val_loss: 2.0563 - val_acc: 0.3636
Epoch 6/30
4/3 [========================================] - 3s 702ms/step - loss: 1.8335 - acc: 0.4747 - val_loss: 1.7684 - val_acc: 0.6000
Epoch 7/30
4/3 [========================================] - 2s 607ms/step - loss: 1.6105 - acc: 0.5661 - val_loss: 1.5527 - val_acc: 0.5091
Epoch 8/30
4/3 [========================================] - 2s 608ms/step - loss: 1.2896 - acc: 0.5948 - val_loss: 1.3767 - val_acc: 0.5636
Epoch 9/30
4/3 [========================================] - 2s 620ms/step - loss: 1.2371 - acc: 0.5931 - val_loss: 1.2162 - val_acc: 0.6364
Epoch 10/30
4/3 [========================================] - 2s 594ms/step - loss: 1.2278 - acc: 0.5733 - val_loss: 1.1651 - val_acc: 0.6727
Epoch 11/30
4/3 [========================================] - 2s 613ms/step - loss: 1.0793 - acc: 0.5442 - val_loss: 1.0447 - val_acc: 0.6909
Epoch 12/30
4/3 [========================================] - 2s 599ms/step - loss: 1.0326 - acc: 0.6948 - val_loss: 1.2848 - val_acc: 0.6182
Epoch 13/30
4/3 [========================================] - 2s 600ms/step - loss: 0.8976 - acc: 0.7161 - val_loss: 1.1182 - val_acc: 0.7091
Epoch 14/30
4/3 [========================================] - 2s 602ms/step - loss: 0.8456 - acc: 0.7082 - val_loss: 0.9046 - val_acc: 0.7273
Epoch 15/30
4/3 [========================================] - 3s 753ms/step - loss: 0.7381 - acc: 0.7822 - val_loss: 0.8139 - val_acc: 0.7636
Epoch 16/30
4/3 [========================================] - 3s 775ms/step - loss: 0.5814 - acc: 0.8190 - val_loss: 0.8724 - val_acc: 0.7455
Epoch 17/30
4/3 [========================================] - 2s 616ms/step - loss: 0.6171 - acc: 0.7670 - val_loss: 0.7576 - val_acc: 0.7455
Epoch 18/30
4/3 [========================================] - 3s 667ms/step - loss: 0.7284 - acc: 0.7728 - val_loss: 0.5410 - val_acc: 0.8909
Epoch 19/30
4/3 [========================================] - 3s 630ms/step - loss: 0.5744 - acc: 0.8396 - val_loss: 0.6037 - val_acc: 0.8727
Epoch 20/30
4/3 [========================================] - 2s 617ms/step - loss: 0.4618 - acc: 0.8837 - val_loss: 0.6635 - val_acc: 0.8364
Epoch 21/30
4/3 [========================================] - 3s 684ms/step - loss: 0.4439 - acc: 0.8225 - val_loss: 0.5555 - val_acc: 0.8545
Epoch 22/30
4/3 [========================================] - 2s 616ms/step - loss: 0.3823 - acc: 0.8843 - val_loss: 0.6592 - val_acc: 0.8364
Epoch 23/30
4/3 [========================================] - 2s 615ms/step - loss: 0.4111 - acc: 0.8575 - val_loss: 0.7290 - val_acc: 0.8364
Epoch 24/30
4/3 [========================================] - 2s 613ms/step - loss: 0.4160 - acc: 0.8541 - val_loss: 0.5520 - val_acc: 0.8545
Epoch 25/30
4/3 [========================================] - 2s 619ms/step - loss: 0.3284 - acc: 0.8714 - val_loss: 0.5246 - val_acc: 0.8545
Epoch 26/30
4/3 [========================================] - 2s 609ms/step - loss: 0.2572 - acc: 0.8955 - val_loss: 0.6713 - val_acc: 0.8545
Epoch 27/30
4/3 [========================================] - 3s 682ms/step - loss: 0.3457 - acc: 0.8719 - val_loss: 0.5676 - val_acc: 0.8727
Epoch 28/30
4/3 [========================================] - 3s 646ms/step - loss: 0.3611 - acc: 0.8961 - val_loss: 0.4857 - val_acc: 0.9091
Epoch 29/30
4/3 [========================================] - 2s 622ms/step - loss: 0.2753 - acc: 0.9229 - val_loss: 0.5467 - val_acc: 0.8727
Epoch 30/30
4/3 [========================================] - 3s 662ms/step - loss: 0.3308 - acc: 0.8648 - val_loss: 0.5625 - val_acc: 0.8727
Model saved!

print(history.history.keys())

dict_keys(['acc', 'val_acc', 'val_loss', 'loss'])

# summarize history for accuracy
plt.plot(history.history['acc'])
plt.plot(history.history['val_acc'])
plt.title('model accuracy')
plt.ylabel('accuracy')
plt.xlabel('epoch')
plt.legend(['train', 'test'], loc='upper left')
plt.show()

# summarize history for loss
plt.plot(history.history['loss'])
plt.plot(history.history['val_loss'])
plt.title('model loss')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(['train', 'test'], loc='upper left')
plt.show()

alt text