神经网络之AlexNet模型的实现（Python+TensorFlow）

# -*- coding:utf-8 -*-
#
# AlexNet model

import tensorflow as tf
import numpy as np

FLAGS = tf.app.flags.FLAGS
tf.app.flags.DEFINE_integer('num_class', 10, "number of classes.")

#显示网络每一层结构的函数
def print_layer(t):
print(t.op.name, ' ', t.get_shape().as_list())

def alexnet(images):
# conv1
with tf.name_scope('conv1') as scope:
kernel = tf.Variable(tf.truncated_normal([11,11,3,64], stddev=0.1), name='weights')
biases = tf.Variable(tf.constant(0.0, shape=[64], trainable=True, name='biases')
conv_plus_b = tf.nn.bias_add(tf.nn.conv2d(images, kernel, [1,4,4,1], padding='SAME'), biases)
conv1 = tf.nn.relu(conv_plus_b, name=scope)
print_layer(conv1)

# LRN层, pool层
lrn1 = tf.nn.lrn(conv1, 4, bias=1.0, alpha=0.001/9, beta=0.75, name='lrn1')
pool1 = tf.nn.max_pool(lrn1, ksize=[1,3,3,1], strides=[1,2,2,1], padding='VALID', name='pool1')

# conv2
with tf.name_scope('conv2') as scope:
kernel = tf.Variable(tf.truncated_normal([5,5,64,192], stddev=0.1), name='weights')
biases = tf.Variable(tf.constant(0.0, shape=[192]), trainable=True, name='biases')
conv_plus_b = tf.nn.bias_add(tf.nn.conv2d(pool1, kernel, [1,1,1,1], padding='SAME'), biases)
conv2 = tf.nn.relu(conv_plus_b, name=scope)

# LRN层, pool层
lrn2 = tf.nn.lrn(conv2, 4, bias=1.0, alpha=0.001/9, beta=0.75, name='lrn2')
pool2 = tf.nn.max_pool(lrn2, ksize=[1,3,3,1], strides=[1,2,2,1], padding='VALID', name='pool2')

# conv3
with tf.name_scope('conv3') as scope:
kernel = tf.Variable(tf.truncated_normal([3,3,192,384], stddev=0.1), name='weights')
biases = tf.Variable(tf.constant(0.0, shape=[384]), trainable=True, name='biases')
conv_plus_b = tf.nn.bias_add(tf.nn.conv2d(pool2, kernel, [1,1,1,1], padding='SAME'), biases)
conv3 = tf.nn.relu(conv_plus_b, name=scope)

# conv4
with tf.name_scope('conv4') as scope:
kernel = tf.Variable(tf.truncated_normal([3,3,384,256], stddev=0.1), name='weights')
biases = tf.Variable(tf.constant(0.0, shape=[256]), trainable=True, name='biases')
conv_plus_b = tf.nn.bias_add(tf.nn.conv2d(conv3, kernel, [1,1,1,1], padding='SAME'), biases)
conv4 = tf.nn.relu(conv_plus_b, name=scope)

# conv5
with tf.name_scope('conv5') as scope:
kernel = tf.Variable(tf.truncated_normal([3,3,256,256], stddev=0.1), name='weights')
biases = tf.Variable(tf.constant(0.0, shape=[256]), trainable=True, name='biases')
conv_plus_b = tf.nn.bias_add(tf.nn.conv2d(conv4, kernel, [1,1,1,1], padding='SAME'), biases)
conv5 = tf.nn.relu(conv_plus_b, name=scope)

# pool5
pool5 = tf.nn.max_pool(conv5, ksize=[1,3,3,1], strides=[1,2,2,1], padding='VALID', name='pool5')

# 卷积部分到此为止,分类问题的话还有3个全连接层: 4096,4096,num_class
with tf.name_scope('fc1') as scope:
input_shape = pool5.get_shape().as_list()
input_units = np.prod(input_shape[1:])
W_fc1 = tf.Variable(tf.truncated_normal([input_units,4096], stddev=0.1, name='weight')
b_fc1 = tf.Variable(tf.constant(0.1, shape=[4096]), name='bias')
fc1 = tf.nn.relu(tf.matmul(tf.reshape(pool5, [-1,input_units]), W_fc1) + b_fc1, name=scope)
fc1_drop = tf.nn.dropout(fc1, keep_prob)

with tf.name_scope('fc2') as scope:
W_fc2 = tf.Variable(tf.truncated_normal([4096,4096], stddev=0.1, name='weight')
b_fc2 = tf.Variable(tf.constant(0.1, shape=[4096]), name='bias')
fc2 = tf.nn.relu(tf.matmul(fc1_drop, W_fc2) + b_fc2, name=scope)
fc2_drop = tf.nn.dropout(fc2, keep_prob)

with tf.name_scope('fc3') as scope:
W_fc3 = tf.Variable(tf.truncated_normal([4096,FLAGS.num_class], stddev=0.1, name='weight')
b_fc3 = tf.Variable(tf.constant(0.1, shape=[FLAGS.num_class]), name='bias')
out = tf.nn.softmax(tf.matmul(fc2_drop, W_fc3) + b_fc3)

return out