def get_iou_callable():
with tf.Graph().as_default(), tf.device("/cpu:0"):
A = tf.placeholder(tf.float32, shape=[None, 4])
B = tf.placehodler(tf.float32, shape=[None, 4])
iou = pairwise_iou(A, B)
sess = tf.Session()
return sess.make_callable(iou, [A, B])
def train():
x = tf.placeholder(dtype=tf.float32,
shape=[None, 224, 224, 3],
name='input')
y = tf.placehodler(dtype=tf.float32,
shape=[None, num_classes],
name='label')
#keep_prob=tf.placehodler(tf.float32)
#建立模型
with slim.arg_scope(vgg_arg_scope()):
outputs, end_points = vgg_16(x, num_classes)
#定义变量
loss = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits(logits=outputs, label=y))
train_step = tf.train.GradientDescentOptimizer(
learning_rate=lr).minimize(loss)
accuracy = tf.reduce_mean(
tf.cast(tf.equal(tf.argmax(outputs, 1), tf.argmax(y, 1)), tf.float32))
#读取batch
images, labels = read_and_decode('./train.tfrecords')
img_batch, label_batch = tf.train.shuffle_batch([images, labels],
batch_size=batch_size,
capacity=392,
min_after_dequeue=200)
label_batch = tf.one_hot(label_batch, num_classes, 1, 0)
init = tf.global_variables_initializer
saver = tf.train.Saver()
with tf.Session() as sess:
sess.run(init)
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
for i in range(max_steps):
batch_x, batch_y = sess.run(img_batch, label_batch)
_, loss_val = sess.run([train_step, loss],
feed_dict={
x: batch_x,
y: batch_y
})
#每10step 打印
if i % 10 == 0:
train_arr = accuracy.eval(feed_dict={x: batch_x, y: batch_y})
print("%s:step [%d] Loss: %f,training accuracy : %g" %
(datetime.now(), i, loss_val, train_arr))
if (i + 1) == max_steps:
saver.save(sess, './model/model.ckpt', global_step=i)
coord.request_stop()
coord.join(threads)
def _build_RNN(self):
with tf.variable_scope('inputs'):
self._xs = tf.placeholder(
tf.float32,
[self._batch_size, self._time_steps, self._input_size],
name='xs')
self._ys = tf.placehodler(
tf.float32,
[self._batch_size, self._time_steps, self._output_size],
name='ys')
with tf.name_scope('RNN'):
l_in_x = tf.reshape(self._xs, [-1, self._input_size], name='2_2D')
#Ws: (in_size, cell_size)
Wi = self._weight_variable([self._input_size, self._cell_size])
print(Wi.name)
#bs: (cell_size, )
bi = self._bias_variable([
self._cell_size,
])
#l_in_y: (batch_size*time_steps, cell_size)
with tf.name_scope('Wx_plus_b'):
l_in_y = tf.matmul(l_in_x, Wi) + bi
l_in_y = tf.reshape(l_in_y,
[-1, self._time_steps, self._cell_size])
with tf.variable_scope('cell'):
cell = tf.contrib.rnn.BasicLSTMCell(self._cell_size)
with tf.name_scope('initial_state'):
self._cell_initial_state = cell.zero_state(self._batch_size,
dtype=tf.float32)
self.cell_outputs = []
cell_state = self._cell_initial_state
for t in range(self._time_steps):
if t > 0:
tf.get_variable_scope().reuse_variables() #reuse Wi and bi
cell_output, cell_state = cell(l_in_y[:, t, :], cell_state)
self.cell_outputs.append(
cell_output
) #cell_outputsL (time_steps, batch_size, output_size)
self._cell_final_state = cell_state
with tf.variable_scope('output_layer'):
#cell_outputs_reshaped (batch_size*time_steps, cell_size)
cell_outputs_reshaped = tf.reshape(
tf.transpose(self.cell_outputs, [1, 0, 2]),
[-1, self._cell_size])
"""
import cifar10, cifar10_input
import tensorflow as tf
import numpy as np
import time
"""1.定义参数和算法公式"""
training_nums = 3000
batch_size = 128
data_dir = 'xxx?oooxoxoxox'
cifar10.maybe_download_and_extract()
train_images, train_labels = cifar10_input.distorted_inputs(
data_dir=data_dir, batch_size=batch_size)
test_images, test_labels = cifar10_input.inputs(eval_data=True,
data_dir=data_dir,
batch_size=batch_size)
images_ph = tf.placehodler(tf.float32, [batch_size, 24, 24, 3])
labels_ph = tf.placehodler(tf.int32, [batch_size])
"""2.定义权重"""
def weight_variable(shape, stddev, w1):
var = tf.Variable(tf.truncated_normal(shape=shape, stddev=stddev))
if w1 is not None: #对权重做L2正则化
weight_loss = tf.multiply(tf.nn.l2_loss(var), w1, name='weight_loss')
tf.add_to_collection('losses', weight_loss)
return var
"""3.定义偏置"""
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("/tmp/data", one_hot=True)
# [0,0,1,0,0,0,0,0,0,0] = 2
n_classes = 10
batch_size = 128
x = tf.placehodler('float', [None, 784])
y = tf.placeholder('float')
def conv2d(x, W):
return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding="SAME")
def maxpool2d(x):
return tf.nn.max_pool(x,
ksize=[1, 2, 2, 1],
strides=[1, 2, 2, 1],
padding="SAME")
def cnn(x):
weight = {
"W_conv1": tf.Variable(tf.random_normal([5, 5, 1, 32])),
"W_conv2": tf.Variable(tf.random_normal([5, 5, 32, 64])),
"W_fc": tf.Variable(tf.random_normal([7 * 7 * 64, 1024])),
"out": tf.Variable(tf.random_normal([1024, n_classes]))
def add_placeholders(self):
self.input_placeholder = tf.placehodler(tf.init32, [None, self.config.window_size])
self.labels_placeholder = tf.placeholder(tf.float32, [None, self.config.label_size])