一、前期工作
本文将实现灵笼中人物角色的识别。较上一篇文章,这次我采用了VGG-19结构,并增加了预测与保存and加载模型两个部分。
我的环境:
- 语言环境:Python3.6.5
- 编译器:jupyter notebook
- 深度学习环境:TensorFlow2
往期精彩内容:
- 深度学习100例-卷积神经网络(CNN)实现mnist手写数字识别 | 第1天
- 深度学习100例-卷积神经网络(CNN)彩色图片分类 | 第2天
- 深度学习100例-卷积神经网络(CNN)服装图像分类 | 第3天
- 深度学习100例-卷积神经网络(CNN)花朵识别 | 第4天
- 深度学习100例-卷积神经网络(CNN)天气识别 | 第5天
- 深度学习100例-卷积神经网络(VGG-16)识别海贼王草帽一伙 | 第6天
来自专栏:【深度学习100例】
1. 设置GPU
如果使用的是CPU可以忽略这步
代码语言:javascript复制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
代码语言:javascript复制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
# 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)
from tensorflow import keras
from tensorflow.keras import layers,models
import pathlib
代码语言:javascript复制data_dir = "D:/jupyter notebook/DL-100-days/datasets/linglong_photos"
data_dir = pathlib.Path(data_dir)
3. 查看数据
数据集中一共有白月魁、查尔斯、红蔻、马克、摩根、冉冰等6个人物角色。
文件夹 | 含义 | 数量 |
---|---|---|
baiyuekui | 白月魁 | 40 张 |
chaersi | 查尔斯 | 76 张 |
hongkou | 红蔻 | 36 张 |
make | 马克 | 38张 |
mogen | 摩根 | 30 张 |
ranbing | 冉冰 | 60张 |
image_count = len(list(data_dir.glob('*/*')))
print("图片总数为:",image_count)
代码语言:javascript复制图片总数为: 280
二、数据预处理
1. 加载数据
使用image_dataset_from_directory
方法将磁盘中的数据加载到tf.data.Dataset
中
batch_size = 16
img_height = 224
img_width = 224
代码语言:javascript复制"""
关于image_dataset_from_directory()的详细介绍可以参考文章:https://mtyjkh.blog.csdn.net/article/details/117018789
"""
train_ds = tf.keras.preprocessing.image_dataset_from_directory(
data_dir,
validation_split=0.1,
subset="training",
seed=123,
image_size=(img_height, img_width),
batch_size=batch_size)
代码语言:javascript复制Found 280 files belonging to 6 classes.
Using 252 files for training.
代码语言:javascript复制"""
关于image_dataset_from_directory()的详细介绍可以参考文章:https://mtyjkh.blog.csdn.net/article/details/117018789
"""
val_ds = tf.keras.preprocessing.image_dataset_from_directory(
data_dir,
validation_split=0.1,
subset="validation",
seed=123,
image_size=(img_height, img_width),
batch_size=batch_size)
代码语言:javascript复制Found 280 files belonging to 6 classes.
Using 28 files for validation.
我们可以通过class_names输出数据集的标签。标签将按字母顺序对应于目录名称。
代码语言:javascript复制class_names = train_ds.class_names
print(class_names)
代码语言:javascript复制['baiyuekui', 'chaersi', 'hongkou', 'make', 'mogen', 'ranbing']
2. 可视化数据
代码语言:javascript复制plt.figure(figsize=(10, 5)) # 图形的宽为10高为5
plt.suptitle("微信公众号:K同学啊")
for images, labels in train_ds.take(1):
for i in range(8):
ax = plt.subplot(2, 4, i 1)
plt.imshow(images[i].numpy().astype("uint8"))
plt.title(class_names[labels[i]])
plt.axis("off")
代码语言:javascript复制plt.imshow(images[1].numpy().astype("uint8"))
3. 再次检查数据
代码语言:javascript复制for image_batch, labels_batch in train_ds:
print(image_batch.shape)
print(labels_batch.shape)
break
代码语言:javascript复制(16, 224, 224, 3)
(16,)
Image_batch
是形状的张量(32,180,180,3)。这是一批形状180x180x3的32张图片(最后一维指的是彩色通道RGB)。Label_batch
是形状(32,)的张量,这些标签对应32张图片
4. 配置数据集
- shuffle() : 打乱数据,关于此函数的详细介绍可以参考:https://zhuanlan.zhihu.com/p/42417456
- prefetch() :预取数据,加速运行,其详细介绍可以参考我前两篇文章,里面都有讲解。
- cache() :将数据集缓存到内存当中,加速运行
AUTOTUNE = tf.data.AUTOTUNE
train_ds = train_ds.cache().shuffle(1000).prefetch(buffer_size=AUTOTUNE)
val_ds = val_ds.cache().prefetch(buffer_size=AUTOTUNE)
5. 归一化
代码语言:javascript复制normalization_layer = layers.experimental.preprocessing.Rescaling(1./255)
normalization_train_ds = train_ds.map(lambda x, y: (normalization_layer(x), y))
val_ds = val_ds.map(lambda x, y: (normalization_layer(x), y))
代码语言:javascript复制image_batch, labels_batch = next(iter(val_ds))
first_image = image_batch[0]
# 查看归一化后的数据
print(np.min(first_image), np.max(first_image))
代码语言:javascript复制0.00390696 1.0
三、构建VGG-19网络
在官方模型与自建模型之间进行二选一就可以啦,选着一个注释掉另外一个,都是正版的VGG-19哈。
VGG优缺点分析:
- VGG优点
VGG的结构非常简洁,整个网络都使用了同样大小的卷积核尺寸(3x3)
和最大池化尺寸(2x2)
。
- VGG缺点
1)训练时间过长,调参难度大。2)需要的存储容量大,不利于部署。例如存储VGG-16
权重值文件的大小为500多MB,不利于安装到嵌入式系统中。
1. 官方模型(已打包好)
官网模型调用这块我放到后面几篇文章中,下面主要讲一下VGG-19
代码语言:javascript复制# model = keras.applications.VGG19(weights='imagenet')
# model.summary()
2. 自建模型
代码语言:javascript复制from tensorflow.keras import layers, models, Input
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Conv2D, MaxPooling2D, Dense, Flatten, Dropout
def VGG19(nb_classes, input_shape):
input_tensor = Input(shape=input_shape)
# 1st block
x = Conv2D(64, (3,3), activation='relu', padding='same',name='block1_conv1')(input_tensor)
x = Conv2D(64, (3,3), activation='relu', padding='same',name='block1_conv2')(x)
x = MaxPooling2D((2,2), strides=(2,2), name = 'block1_pool')(x)
# 2nd block
x = Conv2D(128, (3,3), activation='relu', padding='same',name='block2_conv1')(x)
x = Conv2D(128, (3,3), activation='relu', padding='same',name='block2_conv2')(x)
x = MaxPooling2D((2,2), strides=(2,2), name = 'block2_pool')(x)
# 3rd block
x = Conv2D(256, (3,3), activation='relu', padding='same',name='block3_conv1')(x)
x = Conv2D(256, (3,3), activation='relu', padding='same',name='block3_conv2')(x)
x = Conv2D(256, (3,3), activation='relu', padding='same',name='block3_conv3')(x)
x = Conv2D(256, (3,3), activation='relu', padding='same',name='block3_conv4')(x)
x = MaxPooling2D((2,2), strides=(2,2), name = 'block3_pool')(x)
# 4th block
x = Conv2D(512, (3,3), activation='relu', padding='same',name='block4_conv1')(x)
x = Conv2D(512, (3,3), activation='relu', padding='same',name='block4_conv2')(x)
x = Conv2D(512, (3,3), activation='relu', padding='same',name='block4_conv3')(x)
x = Conv2D(512, (3,3), activation='relu', padding='same',name='block4_conv4')(x)
x = MaxPooling2D((2,2), strides=(2,2), name = 'block4_pool')(x)
# 5th block
x = Conv2D(512, (3,3), activation='relu', padding='same',name='block5_conv1')(x)
x = Conv2D(512, (3,3), activation='relu', padding='same',name='block5_conv2')(x)
x = Conv2D(512, (3,3), activation='relu', padding='same',name='block5_conv3')(x)
x = Conv2D(512, (3,3), activation='relu', padding='same',name='block5_conv4')(x)
x = MaxPooling2D((2,2), strides=(2,2), name = 'block5_pool')(x)
# full connection
x = Flatten()(x)
x = Dense(4096, activation='relu', name='fc1')(x)
x = Dense(4096, activation='relu', name='fc2')(x)
output_tensor = Dense(nb_classes, activation='softmax', name='predictions')(x)
model = Model(input_tensor, output_tensor)
return model
model=VGG19(1000, (img_width, img_height, 3))
model.summary()
代码语言:javascript复制Model: "model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_1 (InputLayer) [(None, 224, 224, 3)] 0
_________________________________________________________________
block1_conv1 (Conv2D) (None, 224, 224, 64) 1792
_________________________________________________________________
block1_conv2 (Conv2D) (None, 224, 224, 64) 36928
_________________________________________________________________
block1_pool (MaxPooling2D) (None, 112, 112, 64) 0
_________________________________________________________________
block2_conv1 (Conv2D) (None, 112, 112, 128) 73856
_________________________________________________________________
block2_conv2 (Conv2D) (None, 112, 112, 128) 147584
_________________________________________________________________
block2_pool (MaxPooling2D) (None, 56, 56, 128) 0
_________________________________________________________________
block3_conv1 (Conv2D) (None, 56, 56, 256) 295168
_________________________________________________________________
block3_conv2 (Conv2D) (None, 56, 56, 256) 590080
_________________________________________________________________
block3_conv3 (Conv2D) (None, 56, 56, 256) 590080
_________________________________________________________________
block3_conv4 (Conv2D) (None, 56, 56, 256) 590080
_________________________________________________________________
block3_pool (MaxPooling2D) (None, 28, 28, 256) 0
_________________________________________________________________
block4_conv1 (Conv2D) (None, 28, 28, 512) 1180160
_________________________________________________________________
block4_conv2 (Conv2D) (None, 28, 28, 512) 2359808
_________________________________________________________________
block4_conv3 (Conv2D) (None, 28, 28, 512) 2359808
_________________________________________________________________
block4_conv4 (Conv2D) (None, 28, 28, 512) 2359808
_________________________________________________________________
block4_pool (MaxPooling2D) (None, 14, 14, 512) 0
_________________________________________________________________
block5_conv1 (Conv2D) (None, 14, 14, 512) 2359808
_________________________________________________________________
block5_conv2 (Conv2D) (None, 14, 14, 512) 2359808
_________________________________________________________________
block5_conv3 (Conv2D) (None, 14, 14, 512) 2359808
_________________________________________________________________
block5_conv4 (Conv2D) (None, 14, 14, 512) 2359808
_________________________________________________________________
block5_pool (MaxPooling2D) (None, 7, 7, 512) 0
_________________________________________________________________
flatten (Flatten) (None, 25088) 0
_________________________________________________________________
fc1 (Dense) (None, 4096) 102764544
_________________________________________________________________
fc2 (Dense) (None, 4096) 16781312
_________________________________________________________________
predictions (Dense) (None, 1000) 4097000
=================================================================
Total params: 143,667,240
Trainable params: 143,667,240
Non-trainable params: 0
_________________________________________________________________
3. 网络结构图
关于卷积计算的相关知识可以参考文章:https://mtyjkh.blog.csdn.net/article/details/114278995
结构说明:
- 16个卷积层(Convolutional Layer),分别用
blockX_convX
表示 - 3个全连接层(Fully connected Layer),分别用
fcX
与predictions
表示 - 5个池化层(Pool layer),分别用
blockX_pool
表示
VGG-19包含了19个隐藏层(16个卷积层和3个全连接层),故称为VGG-19
四、编译
在准备对模型进行训练之前,还需要再对其进行一些设置。以下内容是在模型的编译步骤中添加的:
- 损失函数(loss):用于衡量模型在训练期间的准确率。
- 优化器(optimizer):决定模型如何根据其看到的数据和自身的损失函数进行更新。
- 指标(metrics):用于监控训练和测试步骤。以下示例使用了准确率,即被正确分类的图像的比率。
# 设置优化器,我这里改变了学习率。
opt = tf.keras.optimizers.Nadam(learning_rate=1e-5)
model.compile(optimizer=opt,
loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
metrics=['accuracy'])
五、训练模型
代码语言:javascript复制epochs = 10
history = model.fit(
train_ds,
validation_data=val_ds,
epochs=epochs
)
代码语言:javascript复制Epoch 1/10
16/16 [==============================] - 12s 276ms/step - loss: 5.4474 - accuracy: 0.1501 - val_loss: 6.8601 - val_accuracy: 0.0714
Epoch 2/10
16/16 [==============================] - 2s 133ms/step - loss: 1.7873 - accuracy: 0.3191 - val_loss: 6.8396 - val_accuracy: 0.4643
Epoch 3/10
16/16 [==============================] - 2s 137ms/step - loss: 1.4631 - accuracy: 0.4250 - val_loss: 6.8453 - val_accuracy: 0.5714
Epoch 4/10
16/16 [==============================] - 2s 130ms/step - loss: 1.1500 - accuracy: 0.6090 - val_loss: 6.8554 - val_accuracy: 0.3571
Epoch 5/10
16/16 [==============================] - 2s 130ms/step - loss: 1.0349 - accuracy: 0.6292 - val_loss: 6.8421 - val_accuracy: 0.4643
Epoch 6/10
16/16 [==============================] - 2s 131ms/step - loss: 1.0131 - accuracy: 0.5919 - val_loss: 6.8288 - val_accuracy: 0.5714
Epoch 7/10
16/16 [==============================] - 2s 131ms/step - loss: 0.6961 - accuracy: 0.7776 - val_loss: 6.8388 - val_accuracy: 0.6429
Epoch 8/10
16/16 [==============================] - 2s 130ms/step - loss: 0.3716 - accuracy: 0.8975 - val_loss: 6.8132 - val_accuracy: 0.5714
Epoch 9/10
16/16 [==============================] - 2s 130ms/step - loss: 0.3372 - accuracy: 0.8586 - val_loss: 6.8059 - val_accuracy: 0.6071
Epoch 10/10
16/16 [==============================] - 2s 130ms/step - loss: 0.1256 - accuracy: 0.9736 - val_loss: 6.7767 - val_accuracy: 0.8929
六、模型评估
代码语言:javascript复制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.suptitle("微信公众号:K同学啊")
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()
为体现原汁原味的VGG-19,本文并未对模型参数进行修改,可依据实际情况修改模型中的相关性参数,适应实际情况以便提升分类效果。
较上一篇文章【学习100例-卷积神经网络(VGG-16)识别海贼王草帽一伙 | 第6天】我做了如下三个改变:
- 将模型从
VGG-16
改为VGG-19
, - 将学习率(learning_rate)从
1e-4
改为了1e-5
- 更换了数据集
是不是仿佛明白了什么呢
不明白也没关系,后面再逐一讲解,这里先给大家一个体验
七、保存and加载模型
这是最简单的模型保存与加载方法哈
代码语言:javascript复制# 保存模型
model.save('model/my_model.h5')
代码语言:javascript复制# 加载模型
new_model = keras.models.load_model('model/my_model.h5')
八、预测
代码语言:javascript复制# 采用加载的模型(new_model)来看预测结果
plt.figure(figsize=(10, 5)) # 图形的宽为10高为5
plt.suptitle("微信公众号:K同学啊")
for images, labels in val_ds.take(1):
for i in range(8):
ax = plt.subplot(2, 4, 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)
# 使用模型预测图片中的人物
predictions = new_model.predict(img_array)
plt.title(class_names[np.argmax(predictions)])
plt.axis("off")
VGG-19这篇文章其实埋下了很多坑,我都非常巧妙的将它隐藏起来了不知道大家有没有发现。大家可以将自己发现的问题在下方留言处进行讨论。对于一个完美主义者,这些不完美看着真的好难受。后面看看能不能专门出几篇文章来讲这些内容。
其他精彩内容:
- 深度学习100例-卷积神经网络(CNN)实现mnist手写数字识别 | 第1天
- 深度学习100例-卷积神经网络(CNN)彩色图片分类 | 第2天
- 深度学习100例-卷积神经网络(CNN)服装图像分类 | 第3天
- 深度学习100例-卷积神经网络(CNN)花朵识别 | 第4天
- 深度学习100例-卷积神经网络(CNN)天气识别 | 第5天
- 深度学习100例-卷积神经网络(VGG-16)识别海贼王草帽一伙 | 第6天
《深度学习100例》专栏直达:【传送门】