一、数据集的准备和处理
整个项目的目录如下:
|-01_classification_network
|--data
|---train
|---- XX_X.jpg
...
|---test
|---- XX_X.jpg
...
|--logs
|--input_data.py
|--model.py
|--test_all_image.py
|--test_one_image.py
|--training.py
1、数据集的准备
首先在 data 文件夹里创建一个 train 文件夹, train 文件夹里存放要训练的数据集,数据集的命名方式用的是 “类别_排序.jpg的方式”,也可以按照你自己的要求来命名。
2、创建加载数据的程序,并测试
在 data 文件夹的同目录下创建 input_data.py 文件。
1、加载图片
2、获取batch
3、测试下获取batch成功没(测试完注释掉)
import tensorflow as tf
import numpy as np
import os
import matplotlib.pyplot as plt
#设置训练集所在的路径
train_dir = 'C:/Users/huan4/Desktop/Python/TensorFlow/train_ourself_data/01_classification_network/data/train/'
def get_files(file_dir):
'''
Args:
file_dir: file directory
Returns:
list of images and labels
'''
#设置每个类别的图片和标签数组
airplane = []
label_airplane = []
car = []
label_car = []
bird = []
label_bird = []
for file in os.listdir(file_dir):
name = file.split(sep='_')
if name[0]=='airplane':
airplane.append(file_dir + file)
label_airplane.append(0)
elif name[0] == 'car':
car.append(file_dir + file)
label_car.append(1)
elif name[0] == 'bird':
bird.append(file_dir + file)
label_bird.append(2)
print('There are %d airplane\nThere are %d car\nThere are %d bird' %(len(airplane), len(car), len(bird))) #输出每个种类图片的数量
image_list = np.hstack((airplane, car, bird)) #np.hstack 水平拼接这些数组
label_list = np.hstack((label_airplane, label_car, label_bird))
#使用numpy打乱图片和标签的顺序
temp = np.array([image_list, label_list])
temp = temp.transpose()
np.random.shuffle(temp)
#再分别把图片和标签取出来,同时把标签int()化
image_list = list(temp[:, 0])
label_list = list(temp[:, 1])
label_list = [int(i) for i in label_list]
return image_list, label_list
def get_batch(image, label, image_W, image_H, batch_size, capacity):
'''
Args:
image: list type
label: list type
image_W: image width
image_H: image height
batch_size: batch size
capacity: the maximum elements in queue
Returns:
image_batch: 4D tensor [batch_size, width, height, 3], dtype=tf.float32
label_batch: 1D tensor [batch_size], dtype=tf.int32
'''
#将numpy的数据格式转成 tf 的数据格式
image = tf.cast(image, tf.string)
label = tf.cast(label, tf.int32)
# 生成输入的队列
input_queue = tf.train.slice_input_producer([image, label])
label = input_queue[1]
image_contents = tf.read_file(input_queue[0])
image = tf.image.decode_jpeg(image_contents, channels=3) #将队列里的图片解码
######################################
# data argumentation should go to here 数据增强的可以加在这里
######################################
#将图片resize为相同大小
image = tf.image.resize_image_with_crop_or_pad(image, image_W, image_H)
#图片归一化
image = tf.image.per_image_standardization(image)
image_batch, label_batch = tf.train.batch([image, label],
batch_size= batch_size,
num_threads= 64,
capacity = capacity)
#you can also use shuffle_batch 打乱数据的顺序,因为前面打乱了,这里就不打乱了
# image_batch, label_batch = tf.train.shuffle_batch([image,label],
# batch_size=BATCH_SIZE,
# num_threads=64,
# capacity=CAPACITY,
# min_after_dequeue=CAPACITY-1)
label_batch = tf.reshape(label_batch, [batch_size])
image_batch = tf.cast(image_batch, tf.float32)
return image_batch, label_batch
def test_batch():
BATCH_SIZE = 2
CAPACITY = 256
IMG_W = 32
IMG_H = 32
image_list, label_list = get_files(train_dir)
image_batch, label_batch = get_batch(image_list, label_list, IMG_W, IMG_H, BATCH_SIZE, CAPACITY)
with tf.Session() as sess:
i = 0 #只看看第一个batch就好
coord = tf.train.Coordinator() #读取batch的操作,入列和出列
threads = tf.train.start_queue_runners(coord=coord)
try:
while not coord.should_stop() and i<1:
img, label = sess.run([image_batch, label_batch])
# just test one batch
for j in np.arange(BATCH_SIZE):
print('label: %d' %label[j])
plt.imshow(img[j,:,:,:])
plt.show()
i+=1
except tf.errors.OutOfRangeError:
print('done!')
finally:
coord.request_stop()
coord.join(threads)
#test_batch() #测试的时候用,训练的时候注释掉
测试效果:
这是归一化的图片,想看原图的话在get_batch()
函数里把图片归一化注释掉image = tf.image.per_image_standardization(image)
,但是训练的时候必须用归一化,这里注释只是为了显示原图。
原图:
二、分类模型的建立
在 data 文件夹的同目录下创建 model.py 文件。
1、 定义网络的结构
conv1 - pool1 - conv2 - pool2 - fc3 - fc4 - softmax
2、 计算 loss
3、 训练,设置学习率,使 loss 降低
4、 评估准确率
import tensorflow as tf
def inference(images, batch_size, n_classes):
'''Build the model
Args:
images: image batch, 4D tensor, tf.float32, [batch_size, width, height, channels]
Returns:
output tensor with the computed logits, float, [batch_size, n_classes]
'''
with tf.variable_scope('conv1') as scope:
weights = tf.get_variable('weights',
shape = [3,3,3, 16], #shape = [kernel size, kernel size, channels(图片的通道数), kernel numbers(卷积核的数量)]
dtype = tf.float32,
initializer=tf.truncated_normal_initializer(stddev=0.1,dtype=tf.float32)) #初始化
biases = tf.get_variable('biases',
shape=[16], #偏置值,和这层的卷积核数量相同
dtype=tf.float32,
initializer=tf.constant_initializer(0.1)) #初始化
conv = tf.nn.conv2d(images, weights, strides=[1,1,1,1], padding='SAME')
pre_activation = tf.nn.bias_add(conv, biases)
conv1 = tf.nn.relu(pre_activation, name= scope.name)
#pool1 and norm1
with tf.variable_scope('pooling1_lrn') as scope:
pool1 = tf.nn.max_pool(conv1, ksize=[1,3,3,1],strides=[1,2,2,1],
padding='SAME', name='pooling1')
norm1 = tf.nn.lrn(pool1, depth_radius=4, bias=1.0, alpha=0.001/9.0,
beta=0.75,name='norm1')
#conv2
with tf.variable_scope('conv2') as scope:
weights = tf.get_variable('weights',
shape=[3,3,16,16],
dtype=tf.float32,
initializer=tf.truncated_normal_initializer(stddev=0.1,dtype=tf.float32))
biases = tf.get_variable('biases',
shape=[16],
dtype=tf.float32,
initializer=tf.constant_initializer(0.1))
conv = tf.nn.conv2d(norm1, weights, strides=[1,1,1,1],padding='SAME')
pre_activation = tf.nn.bias_add(conv, biases)
conv2 = tf.nn.relu(pre_activation, name='conv2')
#pool2 and norm2
with tf.variable_scope('pooling2_lrn') as scope:
norm2 = tf.nn.lrn(conv2, depth_radius=4, bias=1.0, alpha=0.001/9.0,
beta=0.75,name='norm2')
pool2 = tf.nn.max_pool(norm2, ksize=[1,3,3,1], strides=[1,1,1,1],
padding='SAME',name='pooling2')
#local3
with tf.variable_scope('local3') as scope:
reshape = tf.reshape(pool2, shape=[batch_size, -1])
dim = reshape.get_shape()[1].value
weights = tf.get_variable('weights',
shape=[dim,128],
dtype=tf.float32,
initializer=tf.truncated_normal_initializer(stddev=0.005,dtype=tf.float32))
biases = tf.get_variable('biases',
shape=[128],
dtype=tf.float32,
initializer=tf.constant_initializer(0.1))
local3 = tf.nn.relu(tf.matmul(reshape, weights) + biases, name=scope.name)
#local4
with tf.variable_scope('local4') as scope:
weights = tf.get_variable('weights',
shape=[128,128],
dtype=tf.float32,
initializer=tf.truncated_normal_initializer(stddev=0.005,dtype=tf.float32))
biases = tf.get_variable('biases',
shape=[128],
dtype=tf.float32,
initializer=tf.constant_initializer(0.1))
local4 = tf.nn.relu(tf.matmul(local3, weights) + biases, name='local4')
# softmax
with tf.variable_scope('softmax_linear') as scope:
weights = tf.get_variable('softmax_linear',
shape=[128, n_classes],
dtype=tf.float32,
initializer=tf.truncated_normal_initializer(stddev=0.005,dtype=tf.float32))
biases = tf.get_variable('biases',
shape=[n_classes],
dtype=tf.float32,
initializer=tf.constant_initializer(0.1))
softmax_linear = tf.add(tf.matmul(local4, weights), biases, name='softmax_linear')
return softmax_linear
#%%
def losses(logits, labels):
'''Compute loss from logits and labels
Args:
logits: logits tensor, float, [batch_size, n_classes]
labels: label tensor, tf.int32, [batch_size]
Returns:
loss tensor of float type
'''
with tf.variable_scope('loss') as scope:
cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits\
(logits=logits, labels=labels, name='xentropy_per_example')
loss = tf.reduce_mean(cross_entropy, name='loss')
tf.summary.scalar(scope.name+'/loss', loss)
return loss
def trainning(loss, learning_rate):
'''Training ops, the Op returned by this function is what must be passed to
'sess.run()' call to cause the model to train.
Args:
loss: loss tensor, from losses()
Returns:
train_op: The op for trainning
'''
with tf.name_scope('optimizer'):
optimizer = tf.train.AdamOptimizer(learning_rate= learning_rate)
global_step = tf.Variable(0, name='global_step', trainable=False)
train_op = optimizer.minimize(loss, global_step= global_step)
return train_op
def evaluation(logits, labels):
"""Evaluate the quality of the logits at predicting the label.
Args:
logits: Logits tensor, float - [batch_size, NUM_CLASSES].
labels: Labels tensor, int32 - [batch_size], with values in the
range [0, NUM_CLASSES).
Returns:
A scalar int32 tensor with the number of examples (out of batch_size)
that were predicted correctly.
"""
with tf.variable_scope('accuracy') as scope:
correct = tf.nn.in_top_k(logits, labels, 1)
correct = tf.cast(correct, tf.float16)
accuracy = tf.reduce_mean(correct)
tf.summary.scalar(scope.name+'/accuracy', accuracy)
return accuracy
三、训练模型
在 data 文件夹的同目录下创建 training.py 文件。
import os
import numpy as np
import tensorflow as tf
import input_data
import model
N_CLASSES = 3 #设置你自己数据集的类别数
IMG_W = 32
IMG_H = 32
BATCH_SIZE = 16
CAPACITY = 2000
MAX_STEP = 10000 # 设置训练的步数,每步训练一个 batch_size
learning_rate = 0.0001 # 学习率的设置
def run_training():
train_dir = 'C:/Users/huan4/Desktop/Python/TensorFlow/train_ourself_data/01_classification_network/data/train/'
logs_train_dir ='C:/Users/huan4/Desktop/Python/TensorFlow/train_ourself_data/01_classification_network/logs/train/'
train, train_label = input_data.get_files(train_dir)
train_batch, train_label_batch = input_data.get_batch(train,
train_label,
IMG_W,
IMG_H,
BATCH_SIZE,
CAPACITY)
train_logits = model.inference(train_batch, BATCH_SIZE, N_CLASSES)
train_loss = model.losses(train_logits, train_label_batch)
train_op = model.trainning(train_loss, learning_rate)
train__acc = model.evaluation(train_logits, train_label_batch)
summary_op = tf.summary.merge_all()
sess = tf.Session()
train_writer = tf.summary.FileWriter(logs_train_dir, sess.graph)
saver = tf.train.Saver()
sess.run(tf.global_variables_initializer())
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
try:
for step in np.arange(MAX_STEP):
if coord.should_stop():
break
_, tra_loss, tra_acc = sess.run([train_op, train_loss, train__acc])
if step % 50 == 0:
print('Step %d, train loss = %.2f, train accuracy = %.2f%%' %(step, tra_loss, tra_acc*100.0))
summary_str = sess.run(summary_op)
train_writer.add_summary(summary_str, step)
if step % 2000 == 0 or (step + 1) == MAX_STEP:
checkpoint_path = os.path.join(logs_train_dir, 'model.ckpt')
saver.save(sess, checkpoint_path, global_step=step)
except tf.errors.OutOfRangeError:
print('Done training -- epoch limit reached')
finally:
coord.request_stop()
coord.join(threads)
sess.close()
run_training()
训练记录
四、训练测试
4.1、测试单张图片
在 data 文件夹的同目录下创建 test_one_image.py 文件。
import numpy as np
import tensorflow as tf
import input_data
import model
from PIL import Image
import matplotlib.pyplot as plt
def get_one_image(train, train_label):
'''Randomly pick one image from training data
Return: ndarray
'''
n = len(train)
ind = np.random.randint(0, n)
img_dir = train[ind]
label = train_label[ind]
image = Image.open(img_dir)
plt.imshow(image)
image = image.resize([32, 32])
image = np.array(image)
return image,label,img_dir
def evaluate_one_image():
train_dir = 'C:/Users/huan4/Desktop/Python/TensorFlow/train_ourself_data/01_classification_network/data/train/'
train, train_label = input_data.get_files(train_dir)
image_array, label, img_dir = get_one_image(train, train_label)
name = img_dir.split(sep='/')
with tf.Graph().as_default():
BATCH_SIZE = 1
N_CLASSES = 3
image = tf.cast(image_array, tf.float32)
image = tf.image.per_image_standardization(image)
image = tf.reshape(image, [1, 32, 32, 3])
logit = model.inference(image, BATCH_SIZE, N_CLASSES)
logit = tf.nn.softmax(logit)
x = tf.placeholder(tf.float32, shape=[32, 32, 3])
logs_train_dir = 'C:/Users/huan4/Desktop/Python/TensorFlow/train_ourself_data/01_classification_network/logs/train/'
saver = tf.train.Saver()
with tf.Session() as sess:
print("Reading checkpoints...")
ckpt = tf.train.get_checkpoint_state(logs_train_dir)
if ckpt and ckpt.model_checkpoint_path:
global_step = ckpt.model_checkpoint_path.split('/')[-1].split('-')[-1]
saver.restore(sess, ckpt.model_checkpoint_path)
print('Loading success, global_step is %s' % global_step)
else:
print('No checkpoint file found')
prediction = sess.run(logit, feed_dict={x: image_array})
max_index = np.argmax(prediction)
print('This image is %s, lable is %s' % (name[-1], label))
if max_index==0:
print('This is a airplane with possibility %.6f' %prediction[:, 0])
elif max_index==1:
print('This is a car with possibility %.6f' %prediction[:, 1])
elif max_index==2:
print('This is a bird with possibility %.6f' % prediction[:, 2])
evaluate_one_image()
测试结果:
4.2、测试所有图片
在 ./data/test 文件夹里存放所有要测试的图片,注意测试图片的命名方式和训练的图片命名方式相同。
在 data 文件夹的同目录下创建 test_all_image.py 文件。
代码:
import tensorflow as tf
import input_data
import model
IMG_W = 32
IMG_H = 32
CAPACITY = 2000
def evaluate_all_image():
test_dir = 'C:/Users/huan4/Desktop/Python/TensorFlow/train_ourself_data/01_classification_network/data/test/'
N_CLASSES = 3
print('-------------------------')
test, test_label = input_data.get_files(test_dir)
BATCH_SIZE = len(test)
print('There are %d test images totally..' % BATCH_SIZE)
print('-------------------------')
test_batch, test_label_batch = input_data.get_batch(test,
test_label,
IMG_W,
IMG_H,
BATCH_SIZE,
CAPACITY)
logits = model.inference(test_batch, BATCH_SIZE, N_CLASSES)
testloss = model.losses(logits, test_label_batch)
testacc = model.evaluation(logits, test_label_batch)
logs_train_dir = 'C:/Users/huan4/Desktop/Python/TensorFlow/train_ourself_data/01_classification_network/logs/train/'
saver = tf.train.Saver()
with tf.Session() as sess:
print("Reading checkpoints...")
ckpt = tf.train.get_checkpoint_state(logs_train_dir)
if ckpt and ckpt.model_checkpoint_path:
global_step = ckpt.model_checkpoint_path.split('/')[-1].split('-')[-1]
saver.restore(sess, ckpt.model_checkpoint_path)
print('Loading success, global_step is %s' % global_step)
else:
print('No checkpoint file found')
print('-------------------------')
coord = tf.train.Coordinator()
threads = tf.train.start_queue_runners(sess=sess, coord=coord)
test_loss,test_acc = sess.run([testloss,testacc])
print('The model\'s loss is %.2f' %test_loss)
correct = int(BATCH_SIZE*test_acc)
print('Correct : %d' % correct)
print('Wrong : %d' % (BATCH_SIZE - correct))
print('The accuracy in test images are %.2f%%' %(test_acc*100.0))
coord.request_stop()
coord.join(threads)
sess.close()
evaluate_all_image()
测试结果: