Chaelin

Chaelin

hello :)

[code] AlexNet

ImageNet Classification with Deep Convolutional Neural Networks 구현Permalink

alexnet

위의 layer 구조를 보고 코드로 구현하였다.

앞서 해석한 논문을 보면 디테일한 설명을 볼 수 있다. ImageNet Classification with Deep Convolutional Neural Networks


import tensorflow as tf
import datetime
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from keras.optimizers import SGD
#데이터셋 설정
train_datagen = ImageDataGenerator(rescale=1./255,width_shift_range=0.2,height_shift_range =0.2,
zoom_range=0.2, horizontal_flip =True, vertical_flip = True)
train_generator = train_datagen.flow_from_directory(
'./catdog/training_set/training_set/',
target_size=(227, 227),
batch_size=16,
class_mode='categorical')
test_datagen = ImageDataGenerator(rescale=1./255,width_shift_range=0.2,height_shift_range =0.2,
zoom_range=0.2, horizontal_flip =True, vertical_flip = True)
test_generator = test_datagen.flow_from_directory(
'./catdog/test_set/test_set/',
target_size=(227, 227),
batch_size=16,
class_mode='categorical')
model = tf.keras.models.Sequential([
# C1 (first layer)
tf.keras.layers.Conv2D(filters=96, kernel_size=(11,11), strides=4, activation='relu', input_shape=(227,227,3)),
tf.keras.layers.BatchNormalization(), # currently use batch normalization instead local response normalization
# overlapping
tf.keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2), padding='valid', data_format=None),
# C2
tf.keras.layers.Conv2D(filters=256, kernel_size=(5,5), strides=1, activation='relu', padding="same"),
tf.keras.layers.BatchNormalization(), # currently use batch normalization instead local response normalization
tf.keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2), padding='valid', data_format=None),
# C3
tf.keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=1, activation='relu',padding="same"),
tf.keras.layers.BatchNormalization(),
# C4
tf.keras.layers.Conv2D(filters=384, kernel_size=(3,3), strides=1, activation='relu',padding="same"),
tf.keras.layers.BatchNormalization(),
# C5
tf.keras.layers.Conv2D(filters=256, kernel_size=(3,3), strides=1, activation='relu',padding="same"),
tf.keras.layers.BatchNormalization(),
tf.keras.layers.MaxPool2D(pool_size=(3,3), strides=(2,2), padding='valid', data_format=None),
# Flatten!
tf.keras.layers.Flatten(),
# F6
tf.keras.layers.Dense(4096, activation='relu'),
tf.keras.layers.Dropout(0.5),
# F7
tf.keras.layers.Dense(4096, activation='relu'),
tf.keras.layers.Dropout(0.5),
# outputlayer, Softmax
tf.keras.layers.Dense(2, activation='softmax')
])
sgd = SGD(lr=0.01,decay=5e-4, momentum=0.9)
model.compile(loss='binary_crossentropy', optimizer='sgd',metrics=['accuracy'])
log_dir="logs/fit/" + datetime.datetime.now().strftime("%Y%m%d-%H%M%S")
tensorboard_callback = tf.keras.callbacks.TensorBoard(log_dir=log_dir)
callback_list = [tensorboard_callback]
model.summary()
# start training
model.fit_generator(train_generator, steps_per_epoch=50, epochs=10)
#모델 저장하기
model.save('my_Alexnet.h5')
#모델 평가하기
print("-------------Evaluate-----------------")
scores = model.evaluate_generator(test_generator,steps=1)
print("%s : %.2f%%" %(model.metrics_names[1],scores[1]*100))
view raw alexnet.py hosted with ❤ by GitHub

  • local response normalization은 요즘 사용하지 않고 대부분 batch normalization을 사용한다고 함.

  • compile 부분을 보면 optimizer을 SGD(stochastic gradient descent)를 사용하였다.

  • 원래는 cifar-10 데이터 셋으로 10개의 클래스를 분류하고자 하였지만 GPU 메모리 부족으로 2개의 클래스를 분류해보았다.


[학습결과]Permalink

accuracy는 56.67%.

cat, dog 이미지라서 accurate가 박살난것일까..

참고Permalink

[1] http://datahacker.rs/tf-alexnet/

[2] https://towardsdatascience.com/implementing-alexnet-cnn-architecture-using-tensorflow-2-0-and-keras-2113e090ad98

[3] https://medium.com/swlh/alexnet-with-tensorflow-46f366559ce8

댓글남기기

참고