Building LeNet Model

Kategori: Deep Learning , 29 Nisan 2020 , JanFranco

---

LeNet, 1998 yılında Yann LeCun tarafından oluşturulmuş ve MNIST veri seti üzerinde çok yüksek başarı sağlamış, basit ve eski bir modeldir.

Görüldüğü gibi 2 Convolution katmanından, 2 Pooling katmanından ve Fully Connected Layerdan oluşmakta. Aynı modeli, MNIST veri seti üzerinden train edelim. İlk olarak mnist veri setini import edelim, yükleyelim ve parçalara bölelim:

import cv2
import numpy as np
from keras.utils import np_utils
from keras.datasets import mnist

(x_train, y_train), (x_test, y_test) = mnist.load_data()

Bir kaç resmi inceleyelim:

for i in range(10):
  random_num = np.random.randint(0, len(x_train))
  random_img = x_train[random_num]
  cv2.imshow("Random Img", random_img)
  cv2.waitKey(0)
  
cv2.destroyAllWindows()

Preprocessing işlemlerini gerçekleştirelim. Resmin boyutunu yeniden düzenleyelim, pixellerin değerlerini float tipine çevirelim, daha sonra da 255'e bölerek 0-1 aralığında tutalım:

img_rows = x_train[0].shape[0]
img_cols = x_train[1].shape[0]

x_train = x_train.reshape(x_train.shape[0], img_rows, img_cols, 1)
x_test = x_test.reshape(x_test.shape[0], img_rows, img_cols, 1)

input_shape = (img_rows, img_cols, 1)

x_train = x_train.astype("float32")
x_test = x_test.astype("float32")

x_train /= 255
x_test /= 255

one-hot-encoding işlemini gerçekleştirelim ve sınıf sayısını alalım:

y_train = np_utils.to_categorical(y_train)
y_test = np_utils.to_categorical(y_test)

num_classes = y_test.shape[1]

Modeli oluşturalım ve compile edelim:

from keras.models import Sequential
from keras.optimizers import Adadelta
from keras.layers import Conv2D, Activation, MaxPooling2D, Flatten, Dense

model = Sequential()

model.add(Conv2D(20, (5,5), padding = 'same', input_shape = input_shape))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size = (2,2), strides = (2,2)))

model.add(Conv2D(50, (5,5), padding = 'same'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size = (2,2), strides = (2,2)))

model.add(Flatten())
model.add(Dense(500))
model.add(Activation('relu'))
model.add(Dense(num_classes))
model.add(Activation('softmax'))

model.compile(loss = 'categorical_crossentropy', optimizer = Adadelta(), metrics = ['accuracy'])

print(model.summary())

>>

Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_1 (Conv2D)            (None, 28, 28, 20)        520       
_________________________________________________________________
activation_1 (Activation)    (None, 28, 28, 20)        0         
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 14, 14, 20)        0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 14, 14, 50)        25050     
_________________________________________________________________
activation_2 (Activation)    (None, 14, 14, 50)        0         
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 7, 7, 50)          0         
_________________________________________________________________
flatten_1 (Flatten)          (None, 2450)              0         
_________________________________________________________________
dense_1 (Dense)              (None, 500)               1225500   
_________________________________________________________________
activation_3 (Activation)    (None, 500)               0         
_________________________________________________________________
dense_2 (Dense)              (None, 10)                5010      
_________________________________________________________________
activation_4 (Activation)    (None, 10)                0         
=================================================================
Total params: 1,256,080
Trainable params: 1,256,080
Non-trainable params: 0
_________________________________________________________________
None

Modeli train edelim:

batch_size = 128
epochs = 10

history = model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs, validation_data=(x_test, y_test), shuffle=True)

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

>>

Train on 60000 samples, validate on 10000 samples
Epoch 1/10
60000/60000 [==============================] - 9s 152us/step - loss: 0.1902 - acc: 0.9408 - val_loss: 0.0467 - val_acc: 0.9857
Epoch 2/10
60000/60000 [==============================] - 3s 53us/step - loss: 0.0457 - acc: 0.9858 - val_loss: 0.0390 - val_acc: 0.9870
Epoch 3/10
60000/60000 [==============================] - 3s 53us/step - loss: 0.0300 - acc: 0.9904 - val_loss: 0.0281 - val_acc: 0.9897
Epoch 4/10
60000/60000 [==============================] - 3s 53us/step - loss: 0.0210 - acc: 0.9933 - val_loss: 0.0234 - val_acc: 0.9919
Epoch 5/10
60000/60000 [==============================] - 3s 54us/step - loss: 0.0163 - acc: 0.9952 - val_loss: 0.0219 - val_acc: 0.9923
Epoch 6/10
60000/60000 [==============================] - 3s 53us/step - loss: 0.0114 - acc: 0.9966 - val_loss: 0.0237 - val_acc: 0.9918
Epoch 7/10
60000/60000 [==============================] - 3s 53us/step - loss: 0.0093 - acc: 0.9973 - val_loss: 0.0262 - val_acc: 0.9921
Epoch 8/10
60000/60000 [==============================] - 3s 53us/step - loss: 0.0065 - acc: 0.9983 - val_loss: 0.0208 - val_acc: 0.9931
Epoch 9/10
60000/60000 [==============================] - 3s 53us/step - loss: 0.0053 - acc: 0.9986 - val_loss: 0.0245 - val_acc: 0.9922
Epoch 10/10
60000/60000 [==============================] - 3s 53us/step - loss: 0.0038 - acc: 0.9989 - val_loss: 0.0255 - val_acc: 0.9922
10000/10000 [==============================] - 0s 47us/step
Test loss: 0.025510146142501435
Test accuracy: 0.9922

Tüm kodlar birlikte:

import cv2
import numpy as np
from keras.utils import np_utils
from keras.datasets import mnist

(x_train, y_train), (x_test, y_test) = mnist.load_data()

for i in range(10):
  random_num = np.random.randint(0, len(x_train))
  random_img = x_train[random_num]
  cv2.imshow("Random Img", random_img)
  cv2.waitKey(0)
  
cv2.destroyAllWindows()

img_rows = x_train[0].shape[0]
img_cols = x_train[1].shape[0]

x_train = x_train.reshape(x_train.shape[0], img_rows, img_cols, 1)
x_test = x_test.reshape(x_test.shape[0], img_rows, img_cols, 1)

input_shape = (img_rows, img_cols, 1)

x_train = x_train.astype("float32")
x_test = x_test.astype("float32")

x_train /= 255
x_test /= 255

y_train = np_utils.to_categorical(y_train)
y_test = np_utils.to_categorical(y_test)

num_classes = y_test.shape[1]

from keras.models import Sequential
from keras.optimizers import Adadelta
from keras.layers import Conv2D, Activation, MaxPooling2D, Flatten, Dense

model = Sequential()

model.add(Conv2D(20, (5,5), padding = 'same', input_shape = input_shape))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size = (2,2), strides = (2,2)))

model.add(Conv2D(50, (5,5), padding = 'same'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size = (2,2), strides = (2,2)))

model.add(Flatten())
model.add(Dense(500))
model.add(Activation('relu'))
model.add(Dense(num_classes))
model.add(Activation('softmax'))

model.compile(loss = 'categorical_crossentropy', optimizer = Adadelta(), metrics = ['accuracy'])

print(model.summary())

batch_size = 128
epochs = 10

history = model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs, validation_data=(x_test, y_test), shuffle=True)

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

---

Sonraki Yazı: Building AlexNet Model

---

Yorumlar