def use(src):
x = tf.placeholder(tf.float32, [
None, train_net.IMAGE_SIZE, train_net.IMAGE_SIZE,
train_net.NUM_CHANNELS
],
name='x-input')
img = Image.open(src) #输入图片路径
img_rgb = img.convert("RGB")
img_rgb = img.resize((227, 227))
data = img_rgb.getdata()
data = np.matrix(data, dtype='float')
xs = tf.reshape(data, [1, 227, 227, 3])
y = inference.alex_net(X=x,
output=2,
dropout=train_net.DROPOUT,
regularizer=None)
#tf.argmax()返回向量中最大值位置,tf.equal()返回两个向量对应位置比较结果 返回值为布尔类型
rel = tf.argmax(y, 1)
variable_averages = tf.train.ExponentialMovingAverage(
train_net.MOVING_AVERAGE_DECAY)
#加载变量的滑动平均值
saver = tf.train.Saver(variable_averages.variables_to_restore())
#
# saver = tf.train.Saver()
cls = 'Male'
with tf.Session() as sess:
#返回模型变量取值的路径
tf.global_variables_initializer().run()
ckpt = tf.train.get_checkpoint_state(train_net.MODEL_SAVE_PATH)
if ckpt and ckpt.model_checkpoint_path:
#ckpt.model_checkpoint_path返回最新的模型变量取值的路径
saver.restore(sess, ckpt.model_checkpoint_path)
global_step = ckpt.model_checkpoint_path.split('/')[-1].split(
'-')[-1]
xs = sess.run(xs)
rel = sess.run(rel, feed_dict={x: xs})
if (rel == 1):
cls = 'Female'
print('result is %s' % (cls))
plt.imshow(img) # 显示图片
plt.axis('on') # 不显示坐标轴
plt.title('%s' % cls)
plt.show()
else:
print('NO checkpoint file found')
return
def evaluate(mnist):
# with tf.Graph().as_default() as g:
# x = tf.placeholder(tf.float32, [None, mnist_inferenceLeNet.INPUT_NODE], name='x-input')
x = tf.placeholder(tf.float32, [
None, train_net.IMAGE_SIZE, train_net.IMAGE_SIZE,
train_net.NUM_CHANNELS
],
name='x-input')
y_ = tf.placeholder(tf.float32, [None, train_net.OUTPUT_NODE],
name='y-input')
reshape_xs = np.reshape(
mnist.validation.images,
(mnist.validation.num_examples, train_net.IMAGE_SIZE,
train_net.IMAGE_SIZE, train_net.NUM_CHANNELS))
validata_feed = {x: reshape_xs, y_: mnist.validation.labels}
y, f = inference.alex_net(X=x,
output=10,
dropout=train_net.DROPOUT,
regularizer=None)
#tf.argmax()返回向量中最大值位置,tf.equal()返回两个向量对应位置比较结果 返回值为布尔类型
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
#数据类型转换
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
# variable_averages = tf.train.ExponentialMovingAverage(train_net.MOVING_AVERAGE_DECAY)
# #加载变量的滑动平均值
# saver = tf.train.Saver(variable_averages.variables_to_restore())
#加载保存模型的变量
saver = tf.train.Saver()
while True:
with tf.Session() as sess:
#返回模型变量取值的路径
ckpt = tf.train.get_checkpoint_state(train_net.MODEL_SAVE_PATH)
if ckpt and ckpt.model_checkpoint_path:
#ckpt.model_checkpoint_path返回最新的模型变量取值的路径
saver.restore(sess, ckpt.model_checkpoint_path)
global_step = ckpt.model_checkpoint_path.split('/')[-1].split(
'-')[-1]
print('前层特征：')
print(sess.run(f, feed_dict=validata_feed))
print(
'After %s traing steps validation accuracy is %g' %
(global_step, sess.run(accuracy, feed_dict=validata_feed)))
else:
print('NO checkpoint file found')
return
time.sleep(EVAL_INTERVAL_SECS)
def evaluate(mnist, i):
x = tf.placeholder(tf.float32, [
1, train_net.IMAGE_SIZE, train_net.IMAGE_SIZE, train_net.NUM_CHANNELS
],
name='x-input')
y, f = inference.alex_net(X=x,
output=10,
dropout=train_net.DROPOUT,
regularizer=None)
result = tf.argmax(y, 1)
variable_averages = tf.train.ExponentialMovingAverage(
train_net.MOVING_AVERAGE_DECAY)
#加载变量的滑动平均值
saver = tf.train.Saver(variable_averages.variables_to_restore())
#加载保存模型的变量
# saver = tf.train.Saver()
with tf.Session() as sess:
#返回模型变量取值的路径
ckpt = tf.train.get_checkpoint_state(train_net.MODEL_SAVE_PATH)
if ckpt and ckpt.model_checkpoint_path:
#ckpt.model_checkpoint_path返回最新的模型变量取值的路径
saver.restore(sess, ckpt.model_checkpoint_path)
global_step = ckpt.model_checkpoint_path.split('/')[-1].split(
'-')[-1]
reshape_xs = np.reshape(
mnist.test.images[i],
(1, train_net.IMAGE_SIZE, train_net.IMAGE_SIZE,
train_net.NUM_CHANNELS))
test_feed = {x: reshape_xs}
result = sess.run(result, feed_dict=test_feed)
print('After %s traing steps test result is %g' %
(global_step, result))
print('truly result is %s' %
(sess.run(tf.argmax(mnist.test.labels[i]))))
else:
print('NO checkpoint file found')
return
def test():
reader = tf.TFRecordReader()
_, serialized_example = reader.read(
filename_queue) # return file_name and file
features = tf.parse_single_example(serialized_example,
features={
'label':
tf.FixedLenFeature([], tf.int64),
'img':
tf.FixedLenFeature([], tf.string),
'width':
tf.FixedLenFeature([], tf.int64),
'height':
tf.FixedLenFeature([], tf.int64),
'channels':
tf.FixedLenFeature([], tf.int64)
}) # return image and label
label = tf.cast(features['label'], tf.float32) # throw label tensor
# height = tf.cast(features['height'],tf.int32)
height = features['height']
width = tf.cast(features['width'], tf.int32)
channels = tf.cast(features['channels'], tf.int32)
img = tf.decode_raw(features['img'], tf.uint8)
print(type(height))
img = tf.reshape(img, [227, 227, 3])
img_batch, label_batch = tf.train.shuffle_batch([img, label],
batch_size=1000,
capacity=5173,
min_after_dequeue=5172,
num_threads=4)
x = tf.placeholder(tf.float32, [
None, train_net.IMAGE_SIZE, train_net.IMAGE_SIZE,
train_net.NUM_CHANNELS
],
name='x-input')
y_ = tf.placeholder(tf.int64, [None], name='y-input')
# validata_feed = {x: reshape_xs, y_: mnist.validation.labels}
y = inference.alex_net(X=x,
output=2,
dropout=train_net.DROPOUT,
regularizer=None)
#tf.argmax()返回向量中最大值位置,tf.equal()返回两个向量对应位置比较结果 返回值为布尔类型
correct_prediction = tf.equal(tf.argmax(y, 1), y_)
#数据类型转换
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
# variable_averages = tf.train.ExponentialMovingAverage(train_net.MOVING_AVERAGE_DECAY)
# #加载变量的滑动平均值
# saver = tf.train.Saver(variable_averages.variables_to_restore())
#加载保存模型的变量
saver = tf.train.Saver()
with tf.Session() as sess:
#返回模型变量取值的路径
tf.global_variables_initializer().run()
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
xs, ys = sess.run([img_batch, label_batch])
ckpt = tf.train.get_checkpoint_state(train_net.MODEL_SAVE_PATH)
if ckpt and ckpt.model_checkpoint_path:
#ckpt.model_checkpoint_path返回最新的模型变量取值的路径
saver.restore(sess, ckpt.model_checkpoint_path)
global_step = ckpt.model_checkpoint_path.split('/')[-1].split(
'-')[-1]
print('test accuracy is %g' %
(sess.run(accuracy, feed_dict={
x: xs,
y_: ys
})))
else:
print('NO checkpoint file found')
return
# Parameters
learn_rate = 0.001
decay_rate = 0.1
batch_size = 64
display_step = 20
n_classes = training.num_labels # we got mad kanji
dropout = 0.8 # Dropout, probability to keep units
imagesize = 227
img_channel = 3
x = tf.placeholder(tf.float32, [None, imagesize, imagesize, img_channel])
y = tf.placeholder(tf.float32, [None, n_classes])
keep_prob = tf.placeholder(tf.float32) # dropout (keep probability)
pred = inference.alex_net(x, keep_prob, n_classes, imagesize, img_channel)
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(pred, y))
global_step = tf.Variable(0, trainable=False)
lr = tf.train.exponential_decay(learn_rate, global_step, 1000, decay_rate, staircase=True)
optimizer = tf.train.AdamOptimizer(learning_rate=lr).minimize(cost, global_step=global_step)
correct_pred = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))
init = tf.initialize_all_variables()
saver = tf.train.Saver()
tf.add_to_collection("x", x)
tf.add_to_collection("y", y)
tf.add_to_collection("keep_prob", keep_prob)
tf.add_to_collection("pred", pred)
# Parameters
learn_rate = 0.001
decay_rate = 0.1
batch_size = 64
display_step = 20
n_classes = training.num_labels # we got mad kanji
dropout = 0.8 # Dropout, probability to keep units
imagesize = 227
img_channel = 3
x = tf.placeholder(tf.float32, [None, imagesize, imagesize, img_channel])
y = tf.placeholder(tf.float32, [None, n_classes])
keep_prob = tf.placeholder(tf.float32) # dropout (keep probability)
pred = inference.alex_net(x, keep_prob, n_classes, imagesize, img_channel)
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(pred, y))
global_step = tf.Variable(0, trainable=False)
lr = tf.train.exponential_decay(learn_rate,
global_step,
1000,
decay_rate,
staircase=True)
optimizer = tf.train.AdamOptimizer(learning_rate=lr).minimize(
cost, global_step=global_step)
correct_pred = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))
init = tf.initialize_all_variables()
def train(mnist):
#定义预输入
x = tf.placeholder(tf.float32,
[None, IMAGE_SIZE, IMAGE_SIZE, NUM_CHANNELS],
name='x-input')
y_ = tf.placeholder(tf.float32, [None, OUTPUT_NODE], name='y-input')
#定义正则化
regularizer = tf.contrib.layers.l2_regularizer(REGULARAZTION_RATE)
#调用神经网计算输出结果
y, _ = inference.alex_net(X=x,
output=OUTPUT_NODE,
dropout=DROPOUT,
regularizer=None)
result = tf.argmax(y, 1, name='out')
#定义统计训练轮数的全局变量
global_step = tf.Variable(0, trainable=False)
#定义滑动平均
variable_averages = tf.train.ExponentialMovingAverage(
MOVING_AVERAGE_DECAY, global_step)
variable_averages_op = variable_averages.apply(tf.trainable_variables())
cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(
logits=y, labels=tf.argmax(y_, 1))
cross_entropy_mean = tf.reduce_mean(cross_entropy)
# loss = cross_entropy_mean+tf.add_n(tf.get_collection('losses'))
loss = cross_entropy_mean
#定义学习率变化
learning_rate = tf.train.exponential_decay(
LEARNING_RATE_BASE, global_step, mnist.train.num_examples / BATCH_SIZE,
LEARNING_RATE_DECAY)
train_step = tf.train.GradientDescentOptimizer(learning_rate).minimize(
loss, global_step=global_step)
# train_step = tf.train.MomentumOptimizer(learning_rate,0.9).minimize(loss,global_step=global_step)
#同时更新滑动平均和网络参数
with tf.control_dependencies([train_step, variable_averages_op]):
train_op = tf.no_op(name='train')
saver = tf.train.Saver()
with tf.Session() as sess:
tf.global_variables_initializer().run()
for i in range(TRAINING_STEPS):
xs, ys = mnist.train.next_batch(BATCH_SIZE)
reshape_xs = np.reshape(
xs, (BATCH_SIZE, IMAGE_SIZE, IMAGE_SIZE, NUM_CHANNELS))
_, loss_value, step = sess.run([train_op, loss, global_step],
feed_dict={
x: reshape_xs,
y_: ys
})
if i % 100 == 0:
print('After %d training steps,loss on training batch is %g' %
(step, loss_value))
#保存checkpoint文件
saver.save(sess,
os.path.join(MODEL_SAVE_PATH, MODEL_NAME),
global_step=global_step)
#保存pb文件
output_graph_def = graph_util.convert_variables_to_constants(
sess, sess.graph_def, ['out'])
with tf.gfile.GFile('model_pb/combined_model.pb', 'wb') as f:
f.write(output_graph_def.SerializeToString())
