tensorflow项目构建流程
博客：http://blog.csdn.net/hjimce
微博：黄锦池-hjimce qq:1393852684
一、构建路线
个人感觉对于任何一个深度学习库，如mxnet、tensorflow、theano、caffe等，基本上我都采用同样的一个学习流程，大体流程如下：
(1)训练阶段：数据打包-》网络构建、训练-》模型保存-》可视化查看损失函数、验证精度
(2)测试阶段：模型加载-》测试图片读取-》预测显示结果
(3)移植阶段：量化、压缩加速-》微调-》C++移植打包-》上线
这边我就以tensorflow为例子，讲解整个流程的大体架构，完成一个深度学习项目所需要熟悉的过程代码。
二、训练、测试阶段
1、tensorflow打包数据
这一步对于tensorflow来说，也可以直接自己在线读取：.jpg图片、标签文件等，然后通过phaceholder变量，把数据送入网络中，进行计算。
不过这种效率比较低，对于大规模训练数据来说，我们需要一个比较高效的方式，tensorflow建议我们采用tfrecoder进行高效数据读取。学习tensorflow一定要学会tfrecoder文件写入、读取，具体示例代码如下：
- #coding=utf-8
- #tensorflow高效数据读取训练
- import tensorflow as tf
- import cv2
- #把train.txt文件格式，每一行：图片路径名 类别标签
- #奖数据打包，转换成tfrecords格式，以便后续高效读取
- def encode_to_tfrecords(lable_file,data_root,new_name='data.tfrecords',resize=None):
- writer=tf.python_io.TFRecordWriter(data_root+'/'+new_name)
- num_example=0
- with open(lable_file,'r') as f:
- for l in f.readlines():
- l=l.split()
- image=cv2.imread(data_root+"/"+l[0])
- if resize is not None:
- image=cv2.resize(image,resize)#为了
- height,width,nchannel=image.shape
- label=int(l[1])
- example=tf.train.Example(features=tf.train.Features(feature={
- 'height':tf.train.Feature(int64_list=tf.train.Int64List(value=[height])),
- 'width':tf.train.Feature(int64_list=tf.train.Int64List(value=[width])),
- 'nchannel':tf.train.Feature(int64_list=tf.train.Int64List(value=[nchannel])),
- 'image':tf.train.Feature(bytes_list=tf.train.BytesList(value=[image.tobytes()])),
- 'label':tf.train.Feature(int64_list=tf.train.Int64List(value=[label]))
- }))
- serialized=example.SerializeToString()
- writer.write(serialized)
- num_example+=1
- print lable_file,"样本数据量：",num_example
- writer.close()
- #读取tfrecords文件
- def decode_from_tfrecords(filename,num_epoch=None):
- filename_queue=tf.train.string_input_producer([filename],num_epochs=num_epoch)#因为有的训练数据过于庞大，被分成了很多个文件，所以第一个参数就是文件列表名参数
- reader=tf.TFRecordReader()
- _,serialized=reader.read(filename_queue)
- example=tf.parse_single_example(serialized,features={
- 'height':tf.FixedLenFeature([],tf.int64),
- 'width':tf.FixedLenFeature([],tf.int64),
- 'nchannel':tf.FixedLenFeature([],tf.int64),
- 'image':tf.FixedLenFeature([],tf.string),
- 'label':tf.FixedLenFeature([],tf.int64)
- })
- label=tf.cast(example['label'], tf.int32)
- image=tf.decode_raw(example['image'],tf.uint8)
- image=tf.reshape(image,tf.pack([
- tf.cast(example['height'], tf.int32),
- tf.cast(example['width'], tf.int32),
- tf.cast(example['nchannel'], tf.int32)]))
- #label=example['label']
- return image,label
- #根据队列流数据格式，解压出一张图片后，输入一张图片，对其做预处理、及样本随机扩充
- def get_batch(image, label, batch_size,crop_size):
- #数据扩充变换
- distorted_image = tf.random_crop(image, [crop_size, crop_size, 3])#随机裁剪
- distorted_image = tf.image.random_flip_up_down(distorted_image)#上下随机翻转
- #distorted_image = tf.image.random_brightness(distorted_image,max_delta=63)#亮度变化
- #distorted_image = tf.image.random_contrast(distorted_image,lower=0.2, upper=1.8)#对比度变化
- #生成batch
- #shuffle_batch的参数：capacity用于定义shuttle的范围，如果是对整个训练数据集，获取batch，那么capacity就应该够大
- #保证数据打的足够乱
- images, label_batch = tf.train.shuffle_batch([distorted_image, label],batch_size=batch_size,
- num_threads=16,capacity=50000,min_after_dequeue=10000)
- #images, label_batch=tf.train.batch([distorted_image, label],batch_size=batch_size)
- # 调试显示
- #tf.image_summary('images', images)
- return images, tf.reshape(label_batch, [batch_size])
- #这个是用于测试阶段，使用的get_batch函数
- def get_test_batch(image, label, batch_size,crop_size):
- #数据扩充变换
- distorted_image=tf.image.central_crop(image,39./45.)
- distorted_image = tf.random_crop(distorted_image, [crop_size, crop_size, 3])#随机裁剪
- images, label_batch=tf.train.batch([distorted_image, label],batch_size=batch_size)
- return images, tf.reshape(label_batch, [batch_size])
- #测试上面的压缩、解压代码
- def test():
- encode_to_tfrecords("data/train.txt","data",(100,100))
- image,label=decode_from_tfrecords('data/data.tfrecords')
- batch_image,batch_label=get_batch(image,label,3)#batch 生成测试
- init=tf.initialize_all_variables()
- with tf.Session() as session:
- session.run(init)
- coord = tf.train.Coordinator()
- threads = tf.train.start_queue_runners(coord=coord)
- for l in range(100000):#每run一次，就会指向下一个样本，一直循环
- #image_np,label_np=session.run([image,label])#每调用run一次，那么
- '''''cv2.imshow("temp",image_np)
- cv2.waitKey()'''
- #print label_np
- #print image_np.shape
- batch_image_np,batch_label_np=session.run([batch_image,batch_label])
- print batch_image_np.shape
- print batch_label_np.shape
- coord.request_stop()#queue需要关闭，否则报错
- coord.join(threads)
- #test()
2、网络架构与训练
经过上面的数据格式处理，接着我们只要写一写网络结构、网络优化方法，把数据搞进网络中就可以了，具体示例代码如下：
- #coding=utf-8
- import tensorflow as tf
- from data_encoder_decoeder import encode_to_tfrecords,decode_from_tfrecords,get_batch,get_test_batch
- import cv2
- import os
- class network(object):
- def __init__(self):
- with tf.variable_scope("weights"):
- self.weights={
- #39*39*3->36*36*20->18*18*20
- 'conv1':tf.get_variable('conv1',[4,4,3,20],initializer=tf.contrib.layers.xavier_initializer_conv2d()),
- #18*18*20->16*16*40->8*8*40
- 'conv2':tf.get_variable('conv2',[3,3,20,40],initializer=tf.contrib.layers.xavier_initializer_conv2d()),
- #8*8*40->6*6*60->3*3*60
- 'conv3':tf.get_variable('conv3',[3,3,40,60],initializer=tf.contrib.layers.xavier_initializer_conv2d()),
- #3*3*60->120
- 'fc1':tf.get_variable('fc1',[3*3*60,120],initializer=tf.contrib.layers.xavier_initializer()),
- #120->6
- 'fc2':tf.get_variable('fc2',[120,6],initializer=tf.contrib.layers.xavier_initializer()),
- }
- with tf.variable_scope("biases"):
- self.biases={
- 'conv1':tf.get_variable('conv1',[20,],initializer=tf.constant_initializer(value=0.0, dtype=tf.float32)),
- 'conv2':tf.get_variable('conv2',[40,],initializer=tf.constant_initializer(value=0.0, dtype=tf.float32)),
- 'conv3':tf.get_variable('conv3',[60,],initializer=tf.constant_initializer(value=0.0, dtype=tf.float32)),
- 'fc1':tf.get_variable('fc1',[120,],initializer=tf.constant_initializer(value=0.0, dtype=tf.float32)),
- 'fc2':tf.get_variable('fc2',[6,],initializer=tf.constant_initializer(value=0.0, dtype=tf.float32))
- }
- def inference(self,images):
- # 向量转为矩阵
- images = tf.reshape(images, shape=[-1, 39,39, 3])# [batch, in_height, in_width, in_channels]
- images=(tf.cast(images,tf.float32)/255.-0.5)*2#归一化处理
- #第一层
- conv1=tf.nn.bias_add(tf.nn.conv2d(images, self.weights['conv1'], strides=[1, 1, 1, 1], padding='VALID'),
- self.biases['conv1'])
- relu1= tf.nn.relu(conv1)
- pool1=tf.nn.max_pool(relu1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')
- #第二层
- conv2=tf.nn.bias_add(tf.nn.conv2d(pool1, self.weights['conv2'], strides=[1, 1, 1, 1], padding='VALID'),
- self.biases['conv2'])
- relu2= tf.nn.relu(conv2)
- pool2=tf.nn.max_pool(relu2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')
- # 第三层
- conv3=tf.nn.bias_add(tf.nn.conv2d(pool2, self.weights['conv3'], strides=[1, 1, 1, 1], padding='VALID'),
- self.biases['conv3'])
- relu3= tf.nn.relu(conv3)
- pool3=tf.nn.max_pool(relu3, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')
- # 全连接层1，先把特征图转为向量
- flatten = tf.reshape(pool3, [-1, self.weights['fc1'].get_shape().as_list()[0]])
- drop1=tf.nn.dropout(flatten,0.5)
- fc1=tf.matmul(drop1, self.weights['fc1'])+self.biases['fc1']
- fc_relu1=tf.nn.relu(fc1)
- fc2=tf.matmul(fc_relu1, self.weights['fc2'])+self.biases['fc2']
- return fc2
- def inference_test(self,images):
- # 向量转为矩阵
- images = tf.reshape(images, shape=[-1, 39,39, 3])# [batch, in_height, in_width, in_channels]
- images=(tf.cast(images,tf.float32)/255.-0.5)*2#归一化处理
- #第一层
- conv1=tf.nn.bias_add(tf.nn.conv2d(images, self.weights['conv1'], strides=[1, 1, 1, 1], padding='VALID'),
- self.biases['conv1'])
- relu1= tf.nn.relu(conv1)
- pool1=tf.nn.max_pool(relu1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')
- #第二层
- conv2=tf.nn.bias_add(tf.nn.conv2d(pool1, self.weights['conv2'], strides=[1, 1, 1, 1], padding='VALID'),
- self.biases['conv2'])
- relu2= tf.nn.relu(conv2)
- pool2=tf.nn.max_pool(relu2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')
- # 第三层
- conv3=tf.nn.bias_add(tf.nn.conv2d(pool2, self.weights['conv3'], strides=[1, 1, 1, 1], padding='VALID'),
- self.biases['conv3'])
- relu3= tf.nn.relu(conv3)
- pool3=tf.nn.max_pool(relu3, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='VALID')
- # 全连接层1，先把特征图转为向量
- flatten = tf.reshape(pool3, [-1, self.weights['fc1'].get_shape().as_list()[0]])
- fc1=tf.matmul(flatten, self.weights['fc1'])+self.biases['fc1']
- fc_relu1=tf.nn.relu(fc1)
- fc2=tf.matmul(fc_relu1, self.weights['fc2'])+self.biases['fc2']
- return fc2
- #计算softmax交叉熵损失函数
- def sorfmax_loss(self,predicts,labels):
- predicts=tf.nn.softmax(predicts)
- labels=tf.one_hot(labels,self.weights['fc2'].get_shape().as_list()[1])
- loss =-tf.reduce_mean(labels * tf.log(predicts))# tf.nn.softmax_cross_entropy_with_logits(predicts, labels)
- self.cost= loss
- return self.cost
- #梯度下降
- def optimer(self,loss,lr=0.001):
- train_optimizer = tf.train.GradientDescentOptimizer(lr).minimize(loss)
- return train_optimizer
- def train():
- encode_to_tfrecords("data/train.txt","data",'train.tfrecords',(45,45))
- image,label=decode_from_tfrecords('data/train.tfrecords')
- batch_image,batch_label=get_batch(image,label,batch_size=50,crop_size=39)#batch 生成测试
- #网络链接,训练所用
- net=network()
- inf=net.inference(batch_image)
- loss=net.sorfmax_loss(inf,batch_label)
- opti=net.optimer(loss)
- #验证集所用
- encode_to_tfrecords("data/val.txt","data",'val.tfrecords',(45,45))
- test_image,test_label=decode_from_tfrecords('data/val.tfrecords',num_epoch=None)
- test_images,test_labels=get_test_batch(test_image,test_label,batch_size=120,crop_size=39)#batch 生成测试
- test_inf=net.inference_test(test_images)
- correct_prediction = tf.equal(tf.cast(tf.argmax(test_inf,1),tf.int32), test_labels)
- accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
- init=tf.initialize_all_variables()
- with tf.Session() as session:
- session.run(init)
- coord = tf.train.Coordinator()
- threads = tf.train.start_queue_runners(coord=coord)
- max_iter=100000
- iter=0
- if os.path.exists(os.path.join("model",'model.ckpt')) is True:
- tf.train.Saver(max_to_keep=None).restore(session, os.path.join("model",'model.ckpt'))
- while iter<max_iter:
- loss_np,_,label_np,image_np,inf_np=session.run([loss,opti,batch_label,batch_image,inf])
- #print image_np.shape
- #cv2.imshow(str(label_np[0]),image_np[0])
- #print label_np[0]
- #cv2.waitKey()
- #print label_np
- if iter%50==0:
- print 'trainloss:',loss_np
- if iter%500==0:
- accuracy_np=session.run([accuracy])
- print '***************test accruacy:',accuracy_np,'*******************'
- tf.train.Saver(max_to_keep=None).save(session, os.path.join('model','model.ckpt'))
- iter+=1
- coord.request_stop()#queue需要关闭，否则报错
- coord.join(threads)
- train()
3、可视化显示
(1)首先再源码中加入需要跟踪的变量：
- <span style="font-size:18px;">tf.scalar_summary("cost_function", loss)#损失函数值</span>
- <span style="font-size:18px;">merged_summary_op = tf.merge_all_summaries()</span>
- <span style="font-size:18px;">summary_writer = tf.train.SummaryWriter('log', session.graph)</span>
(4)然后再session执行的时候，保存：
- summary_str,loss_np,_=session.run([merged_summary_op,loss,opti])
- summary_writer.add_summary(summary_str, iter)
(5)最后只要训练完毕后，直接再终端输入命令：
- python /usr/local/lib/python2.7/dist-packages/tensorflow/tensorboard/tensorboard.py --logdir=log
- <span style="font-size:18px;">http://0.0.0.0:6006</span>
即可观训练曲线。
4、测试阶段
测试阶段主要是直接通过加载图模型、读取参数等，然后直接通过tensorflow的相关函数，进行调用，而不需要网络架构相关的代码；通过内存feed_dict的方式，对相关的输入节点赋予相关的数据，进行前向传导，并获取相关的节点数值。
- #coding=utf-8
- import tensorflow as tf
- import os
- import cv2
- def load_model(session,netmodel_path,param_path):
- new_saver = tf.train.import_meta_graph(netmodel_path)
- new_saver.restore(session, param_path)
- x= tf.get_collection('test_images')[0]#在训练阶段需要调用tf.add_to_collection('test_images',test_images),保存之
- y = tf.get_collection("test_inf")[0]
- batch_size = tf.get_collection("batch_size")[0]
- return x,y,batch_size
- def load_images(data_root):
- filename_queue = tf.train.string_input_producer(data_root)
- image_reader = tf.WholeFileReader()
- key,image_file = image_reader.read(filename_queue)
- image = tf.image.decode_jpeg(image_file)
- return image, key
- def test(data_root="data/race/cropbrown"):
- image_filenames=os.listdir(data_root)
- image_filenames=[(data_root+'/'+i) for i in image_filenames]
- #print cv2.imread(image_filenames[0]).shape
- #image,key=load_images(image_filenames)
- race_listsrc=['black','brown','white','yellow']
- with tf.Session() as session:
- coord = tf.train.Coordinator()
- threads = tf.train.start_queue_runners(coord=coord)
- x,y,batch_size=load_model(session,os.path.join("model",'model_ori_race.ckpt.meta'),
- os.path.join("model",'model_ori_race.ckpt'))
- predict_label=tf.cast(tf.argmax(y,1),tf.int32)
- print x.get_shape()
- for imgf in image_filenames:
- image=cv2.imread(imgf)
- image=cv2.resize(image,(76,76)).reshape((1,76,76,3))
- print "cv shape:",image.shape
- #cv2.imshow("t",image_np[:,:,::-1])
- y_np=session.run(predict_label,feed_dict = {x:image, batch_size:1})
- print race_listsrc[y_np]
- coord.request_stop()#queue需要关闭，否则报错
- coord.join(threads)
4、移植阶段
(1)一个算法经过实验阶段后，接着就要进入移植商用，因此接着需要采用tensorflow的c api函数，直接进行预测推理，首先我们先把tensorflow编译成链接库，然后编写cmake，调用tensorflow链接库：
- <span style="font-size:18px;">bazel build -c opt //tensorflow:libtensorflow.so
- </span>
在bazel-bin/tensorflow目录下会生成libtensorflow.so文件
5、C++ API调用、cmake 编写：
三、熟悉常用API
1、LSTM使用
- <span style="font-size:18px;">import tensorflow.nn.rnn_cell
- lstm = rnn_cell.BasicLSTMCell(lstm_size)#创建一个lstm cell单元类，隐藏层神经元个数为lstm_size
- state = tf.zeros([batch_size, lstm.state_size])#一个序列隐藏层的状态值
- loss = 0.0
- for current_batch_of_words in words_in_dataset:
- output, state = lstm(current_batch_of_words, state)#返回值为隐藏层神经元的输出
- logits = tf.matmul(output, softmax_w) + softmax_b#matmul矩阵点乘
- probabilities = tf.nn.softmax(logits)#softmax输出
- loss += loss_function(probabilities, target_words)</span>
1、one-hot函数：
- <span style="font-size:18px;">#ont hot 可以把训练数据的标签，直接转换成one_hot向量，用于交叉熵损失函数
- import tensorflow as tf
- a=tf.convert_to_tensor([[1],[2],[4]])
- b=tf.one_hot(a,5)</span>
>>b的值为
- <span style="font-size:18px;">[[[ 0. 1. 0. 0. 0.]]
- [[ 0. 0. 1. 0. 0.]]
- [[ 0. 0. 0. 0. 1.]]]</span>
2、assign_sub
- <span style="font-size:18px;">import tensorflow as tf
- x = tf.Variable(10, name="x")
- sub=x.assign_sub(3)#如果直接采用x.assign_sub，那么可以看到x的值也会发生变化
- init_op=tf.initialize_all_variables()
- with tf.Session() as sess:
- sess.run(init_op)
- print sub.eval()
- print x.eval()</span>
- <span style="font-size:18px;">state_ops.assign_sub</span>
也就是说assign函数返回结果值的同时，变量本身的值也会被改变
3、变量查看
- <span style="font-size:18px;"> #查看所有的变量
- for l in tf.all_variables():
- print l.name</span>
4、slice函数：
- <span style="font-size:18px;">import cv2
- import tensorflow as tf
- #slice 函数可以用于切割子矩形图片，参数矩形框的rect,begin=(minx,miny),size=(width,height)
- minx=20
- miny=30
- height=100
- width=200
- image=tf.placeholder(dtype=tf.uint8,shape=(386,386,3))
- rect_image=tf.slice(image,(miny,minx,0),(height,width,-1))
- cvimage=cv2.imread("1.jpg")
- cv2.imshow("cv2",cvimage[miny:(miny+height),minx:(minx+width),:])
- with tf.Session() as sess:
- tfimage=sess.run([rect_image],{image:cvimage})
- cv2.imshow('tf',tfimage[0])
- cv2.waitKey()</span>
5、正太分布随机初始化
- tf.truncated_normal
6、打印操作运算在硬件设备信息
- tf.ConfigProto(log_device_placement=True)
- import tensorflow as tf
- with tf.variable_scope('foo'):#在没有启用reuse的情况下,如果该变量还未被创建,那么就创建该变量,如果已经创建过了,那么就获取该共享变量
- v=tf.get_variable('v',[1])
- with tf.variable_scope('foo',reuse=True):#如果启用了reuse,那么编译的时候,如果get_variable没有遇到一个已经创建的变量,是会出错的
- v1=tf.get_variable('v1',[1])
8、allow_soft_placement的使用：allow_soft_placement=True，允许当在代码中指定tf.device设备，如果设备找不到，那么就采用默认的设备。如果该参数设置为false，当设备找不到的时候，会直接编译不通过。
9、batch normalize调用：
- tf.contrib.layers.batch_norm(x, decay=0.9, updates_collections=None, epsilon=self.epsilon, scale=True
扫一扫
3338
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
