详细讲解用 TensorFlow2 识别验证码
目 录
一、前期工作
1.设置GPU
2.导入数据
3.数据可视化
4.标签数字化
二、构建一个tf.data.Dataset
1.预处理函数
2.加载数据
3.配置数据
三、搭建网络模型
四、编译
五、训练
六、模型评估
七、保存和加载模型
八、预测
一、前期工作
本文将手把手教你用TensorFlow2实现验证的识别，整个项目的完整代码都在文章了哈，大家按顺序copy即可运行。
我的环境：
-
语言环境：Python3.6.5 -
编译器：jupyter notebook -
深度学习环境：TensorFlow2.4.1
来自专栏：《深度学习100例》
1.设置GPU
如果使用的是CPU可以注释掉这部分的代码，不影响运行。
import tensorflow as tf
gpus = tf.config.list_physical_devices("GPU")
if gpus:
tf.config.experimental.set_memory_growth(gpus[0], True) #设置GPU显存用量按需使用
tf.config.set_visible_devices([gpus[0]],"GPU")
2.导入数据
import matplotlib.pyplot as plt
# 支持中文
plt.rcParams['font.sans-serif'] = ['SimHei'] # 用来正常显示中文标签
plt.rcParams['axes.unicode_minus'] = False # 用来正常显示负号
import os,PIL,random,pathlib
# 设置随机种子尽可能使结果可以重现
import numpy as np
np.random.seed(1)
# 设置随机种子尽可能使结果可以重现
import tensorflow as tf
tf.random.set_seed(1)
data_dir = "D:/jupyter notebook/DL-100-days/datasets/captcha"
data_dir = pathlib.Path(data_dir)
all_image_paths = list(data_dir.glob('*'))
all_image_paths = [str(path) for path in all_image_paths]
# 打乱数据
random.shuffle(all_image_paths)
# 获取数据标签
all_label_names = [path.split("\\")[5].split(".")[0] for path in all_image_paths]
image_count = len(all_image_paths)
print("图片总数为：",image_count)
图片总数为：1070
3.数据可视化
plt.figure(figsize=(10,5))
for i in range(20):
plt.subplot(5,4,i+1)
plt.xticks([])
plt.yticks([])
plt.grid(False)
# 显示图片
images = plt.imread(all_image_paths[i])
plt.imshow(images)
# 显示标签
plt.xlabel(all_label_names[i])
plt.show()
4.标签数字化
number = ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9']
alphabet = ['a','b','c','d','e','f','g','h','i','j','k','l','m','n','o','p','q','r','s','t','u','v','w','x','y','z']
char_set = number + alphabet
char_set_len = len(char_set)
label_name_len = len(all_label_names[0])
# 将字符串数字化
def text2vec(text):
vector = np.zeros([label_name_len, char_set_len])
for i, c in enumerate(text):
idx = char_set.index(c)
vector[i][idx] = 1.0
return vector
all_labels = [text2vec(i) for i in all_label_names]
二、构建一个tf.data.Dataset
1.预处理函数
def preprocess_image(image):
image = tf.image.decode_jpeg(image, channels=1)
image = tf.image.resize(image, [50, 200])
return image/255.0
def load_and_preprocess_image(path):
image = tf.io.read_file(path)
return preprocess_image(image)
2.加载数据
构建 tf.data.Dataset 最简单的方法就是使用 from_tensor_slices 方法。
AUTOTUNE = tf.data.experimental.AUTOTUNE
path_ds = tf.data.Dataset.from_tensor_slices(all_image_paths)
image_ds = path_ds.map(load_and_preprocess_image, num_parallel_calls=AUTOTUNE)
label_ds = tf.data.Dataset.from_tensor_slices(all_labels)
image_label_ds = tf.data.Dataset.zip((image_ds, label_ds))
image_label_ds
<ZipDataset shapes: ((50, 200, 1), (5, 36)), types: (tf.float32, tf.float64)>
train_ds = image_label_ds.take(1000) # 前1000个batch
val_ds = image_label_ds.skip(1000) # 跳过前1000，选取后面的
3.配置数据
先复习一下prefetch()函数。prefetch()功能详细介绍：CPU 正在准备数据时，加速器处于空闲状态。相反，当加速器正在训练模型时，CPU 处于空闲状态。因此，训练所用的时间是 CPU 预处理时间和加速器训练时间的总和。prefetch()将训练步骤的预处理和模型执行过程重叠到一起。当加速器正在执行第 N 个训练步时，CPU 正在准备第 N+1 步的数据。这样做不仅可以最大限度地缩短训练的单步用时（而不是总用时），而且可以缩短提取和转换数据所需的时间。如果不使用prefetch()，CPU 和 GPU/TPU 在大部分时间都处于空闲状态：
使用prefetch()可显著减少空闲时间：
BATCH_SIZE = 16
train_ds = train_ds.batch(BATCH_SIZE)
train_ds = train_ds.prefetch(buffer_size=AUTOTUNE)
val_ds = val_ds.batch(BATCH_SIZE)
val_ds = val_ds.prefetch(buffer_size=AUTOTUNE)
val_ds
<PrefetchDataset shapes: ((None, 50, 200, 1), (None, 5, 36)), types: (tf.float32, tf.float64)>
三、搭建网络模型
目前这里主要是带大家跑通代码、整理一下思路，大家可以自行优化网络结构、调整模型参数。后续我也会针对性的出一些调优的案例的。
from tensorflow.keras import datasets, layers, models
model = models.Sequential([
layers.Conv2D(32, (3, 3), activation='relu', input_shape=(50, 200, 1)),#卷积层1，卷积核3*3
layers.MaxPooling2D((2, 2)), #池化层1，2*2采样
layers.Conv2D(64, (3, 3), activation='relu'), #卷积层2，卷积核3*3
layers.MaxPooling2D((2, 2)), #池化层2，2*2采样
layers.Flatten(), #Flatten层，连接卷积层与全连接层
layers.Dense(1000, activation='relu'), #全连接层，特征进一步提取
layers.Dense(label_name_len * char_set_len),
layers.Reshape([label_name_len, char_set_len]),
layers.Softmax() #输出层，输出预期结果
])
# 打印网络结构
model.summary()
Model: "sequential"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d (Conv2D) (None, 48, 198, 32) 320
_________________________________________________________________
max_pooling2d (MaxPooling2D) (None, 24, 99, 32) 0
_________________________________________________________________
conv2d_1 (Conv2D) (None, 22, 97, 64) 18496
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 11, 48, 64) 0
_________________________________________________________________
flatten (Flatten) (None, 33792) 0
_________________________________________________________________
dense (Dense) (None, 1000) 33793000
_________________________________________________________________
dense_1 (Dense) (None, 180) 180180
_________________________________________________________________
reshape (Reshape) (None, 5, 36) 0
_________________________________________________________________
softmax (Softmax) (None, 5, 36) 0
=================================================================
Total params: 33,991,996
Trainable params: 33,991,996
Non-trainable params: 0
_________________________________________________________________
四、编译
model.compile(optimizer="adam",
loss='categorical_crossentropy',
metrics=['accuracy'])
五、训练
epochs = 20
history = model.fit(
train_ds,
validation_data=val_ds,
epochs=epochs
)
Epoch 1/20
63/63 [==============================] - 4s 21ms/step - loss: 3.2998 - accuracy: 0.0934 - val_loss: 2.2876 - val_accuracy: 0.2943
Epoch 2/20
63/63 [==============================] - 1s 9ms/step - loss: 1.7016 - accuracy: 0.5195 - val_loss: 1.2014 - val_accuracy: 0.6314
Epoch 3/20
63/63 [==============================] - 1s 10ms/step - loss: 0.5267 - accuracy: 0.8379 - val_loss: 0.9039 - val_accuracy: 0.7286
Epoch 4/20
63/63 [==============================] - 1s 10ms/step - loss: 0.1911 - accuracy: 0.9442 - val_loss: 0.8609 - val_accuracy: 0.7457
.......
63/63 [==============================] - 1s 10ms/step - loss: 0.0916 - accuracy: 0.9714 - val_loss: 0.8937 - val_accuracy: 0.7886
63/63 [==============================] - 1s 10ms/step - loss: 8.0025e-05 - accuracy: 1.0000 - val_loss: 0.6335 - val_accuracy: 0.8514
Epoch 19/20
63/63 [==============================] - 1s 9ms/step - loss: 6.9294e-05 - accuracy: 1.0000 - val_loss: 0.6396 - val_accuracy: 0.8486
Epoch 20/20
63/63 [==============================] - 1s 10ms/step - loss: 6.0775e-05 - accuracy: 1.0000 - val_loss: 0.6448 - val_accuracy: 0.8486
六、模型评估
acc = history.history['accuracy']
val_acc = history.history['val_accuracy']
loss = history.history['loss']
val_loss = history.history['val_loss']
epochs_range = range(epochs)
plt.figure(figsize=(12, 4))
plt.subplot(1, 2, 1)
plt.plot(epochs_range, acc, label='Training Accuracy')
plt.plot(epochs_range, val_acc, label='Validation Accuracy')
plt.legend(loc='lower right')
plt.title('Training and Validation Accuracy')
plt.subplot(1, 2, 2)
plt.plot(epochs_range, loss, label='Training Loss')
plt.plot(epochs_range, val_loss, label='Validation Loss')
plt.legend(loc='upper right')
plt.title('Training and Validation Loss')
plt.show()
七、保存和加载模型
# 保存模型
model.save('model/12_model.h5')
# 加载模型
new_model = tf.keras.models.load_model('model/12_model.h5')
八、预测
def vec2text(vec):
"""
还原标签（向量->字符串）
"""
text = []
for i, c in enumerate(vec):
text.append(char_set[c])
return "".join(text)
plt.figure(figsize=(10, 8)) # 图形的宽为10高为8
for images, labels in val_ds.take(1):
for i in range(6):
ax = plt.subplot(5, 2, i + 1)
# 显示图片
plt.imshow(images[i])
# 需要给图片增加一个维度
img_array = tf.expand_dims(images[i], 0)
# 使用模型预测验证码
predictions = model.predict(img_array)
plt.title(vec2text(np.argmax(predictions, axis=2)[0]))
plt.axis("off")
可以看到验证码中大部分字符预测都是对的，但是少部分字符还是存在问题，大家可以试试优化一下网络结构，调整网络参数等。本案例适合练习优化技巧，借着这个案例了解一下不同的调整对结果有什么不同。