LeNet实验四分类与四分类变为多个二分类

1. 划分二分类

2. 训练独立的二分类模型

3. 二分类预测结果代码

4. 二分类预测结果

5 改进训练模型

6 优化后预测结果代码

7 优化后预测结果

8 训练四分类模型

9 预测结果代码

10 四分类结果识别

1. 划分二分类

可以根据不同的类别进行多个划分，以实现NonDemented为例，划分为NonDemented和Demented两类，不属于NonDemented的全都属于Demented

2. 训练独立的二分类模型

import tensorflow as tf
from tensorflow.keras import layers, models
from tensorflow.keras.preprocessing.image import ImageDataGenerator

from 文件准备 import data_dir

# 数据生成器
train_datagen = ImageDataGenerator(
    rescale=1./255,
    shear_range=0.2,
    zoom_range=0.2,
    horizontal_flip=True,
    validation_split=0.2  # 20%用于验证
)

train_generator = train_datagen.flow_from_directory(
    data_dir,
    target_size=(28, 28),
    batch_size=32,
    class_mode='binary',
    subset='training'
)

validation_generator = train_datagen.flow_from_directory(
    data_dir,
    target_size=(28, 28),
    batch_size=32,
    class_mode='binary',
    subset='validation'
)

# 构建LeNet-5模型
model = models.Sequential()
model.add(layers.Conv2D(6, (5, 5), activation='relu', input_shape=(28, 28, 3), padding='same'))
model.add(layers.AveragePooling2D((2, 2)))
model.add(layers.Conv2D(16, (5, 5), activation='relu', padding='same'))
model.add(layers.AveragePooling2D((2, 2)))
model.add(layers.Conv2D(120, (5, 5), activation='relu', padding='same'))
model.add(layers.Flatten())
model.add(layers.Dense(84, activation='relu'))
model.add(layers.Dense(1, activation='sigmoid'))

# 编译模型
model.compile(optimizer='adam',
              loss='binary_crossentropy',
              metrics=['accuracy'])

# 训练模型
model.fit(
    train_generator,
    steps_per_epoch=train_generator.samples // train_generator.batch_size,
    epochs=10,
    validation_data=validation_generator,
    validation_steps=validation_generator.samples // validation_generator.batch_size
)

# 保存模型
model.save('lenet_binary_classification_model.h5')

3. 预测结果代码

import tensorflow as tf
from tensorflow.keras.preprocessing import image
import numpy as np
import matplotlib.pyplot as plt

# 加载模型
model = tf.keras.models.load_model('lenet_binary_classification_model.h5')

# 预处理图像
def preprocess_image(img_path):
    img = image.load_img(img_path, target_size=(28, 28))
    img_array = image.img_to_array(img) / 255.0
    img_array = np.expand_dims(img_array, axis=0)
    return img_array

# 预测图像
img_path = 'D:\Pycharm_workspace\LeNet实验_二分类\Demented\moderateDem24.jpg'  # 测试图像路径
img_array = preprocess_image(img_path)
prediction = model.predict(img_array)
predicted_class = 'Demented' if prediction[0][0] > 0.5 else 'NonDemented'

print(f'The predicted class is: {predicted_class}')

# 显示图像
img = image.load_img(img_path, target_size=(28, 28))
plt.imshow(img)
plt.title(f'Predicted: {predicted_class}')
plt.show()

4. 预测结果

Demented结果

NonDemented结果没有。。。。。。

竟然全都没有。。。。因为预测的全部都是Demented

疯狂找原因中

猜测是像素太低使得训练的模型准确率太低

于是重新训练

5 改进训练模型

进行重新训练

import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Conv2D, MaxPooling2D, Flatten, Dense
from tensorflow.keras.preprocessing.image import ImageDataGenerator
import matplotlib.pyplot as plt

# 定义LeNet模型
def create_lenet_model(input_shape):
    model = Sequential([
        Conv2D(6, (5, 5), activation='relu', input_shape=input_shape, padding='same'),
        MaxPooling2D((2, 2), strides=2),
        Conv2D(16, (5, 5), activation='relu'),
        MaxPooling2D((2, 2), strides=2),
        Flatten(),
        Dense(120, activation='relu'),
        Dense(84, activation='relu'),
        Dense(1, activation='sigmoid')
    ])
    model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])
    return model

# 数据增强和数据生成器
train_datagen = ImageDataGenerator(rescale=1./255, validation_split=0.2)

# 训练数据生成器
train_generator = train_datagen.flow_from_directory(
    'D:\Pycharm_workspace\LeNet实验_二分类\image',
    target_size=(176, 208),
    batch_size=32,
    class_mode='binary',
    subset='training'
)

# 验证数据生成器
validation_generator = train_datagen.flow_from_directory(
    'D:\Pycharm_workspace\LeNet实验_二分类\image',
    target_size=(176, 208),
    batch_size=32,
    class_mode='binary',
    subset='validation'
)

# 创建并训练模型
input_shape = (176, 208, 3)
model = create_lenet_model(input_shape)
history = model.fit(train_generator, epochs=10, validation_data=validation_generator)

# 保存模型
model.save('dementia_classification_model.h5')

# 绘制训练和验证损失
plt.plot(history.history['loss'], label='训练损失')
plt.plot(history.history['val_loss'], label='验证损失')
plt.title('训练和验证损失')
plt.xlabel('时期')
plt.ylabel('损失')
plt.legend()
plt.show()

# 绘制训练和验证准确率
plt.plot(history.history['accuracy'], label='训练准确率')
plt.plot(history.history['val_accuracy'], label='验证准确率')
plt.title('训练和验证准确率')
plt.xlabel('时期')
plt.ylabel('准确率')
plt.legend()
plt.show()

这里还有图形画loss与准确率但是我忘记保存了，就用控制台的输出

可以看到loss值非常小而且准确率是100

6 优化后预测结果代码

import tensorflow as tf
from tensorflow.keras.models import load_model
from tensorflow.keras.preprocessing import image
import numpy as np
import matplotlib.pyplot as plt
import os

# 加载模型
model = load_model('dementia_classification_model.h5')

# 定义类别标签
class_labels = ['Demented', 'NonDemented']


# 预测函数
def predict_image(img_path):
    img = image.load_img(img_path, target_size=(176, 208))
    img_array = image.img_to_array(img)
    img_array = np.expand_dims(img_array, axis=0)
    img_array /= 255.0

    prediction = model.predict(img_array)
    predicted_class = class_labels[int(prediction[0] > 0.5)]

    # 显示图像和预测结果
    plt.imshow(image.load_img(img_path))
    plt.title(f'Predicted: {predicted_class}')
    plt.axis('off')
    plt.show()


# 预测并展示结果
img_path = r'D:\Pycharm_workspace\LeNet实验_二分类\image\NonDemented\nonDem1.jpg'  # 替换为你的图片路径
predict_image(img_path)

7 优化后预测结果

图片与预测结果对应上了（右侧是图片链接可以看到是Dem的类型）

NonDem的也是对应上了

就此训练完成

8 训练四分类模型

import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Conv2D, MaxPooling2D, Flatten, Dense
from tensorflow.keras.preprocessing.image import ImageDataGenerator
import matplotlib.pyplot as plt

# 定义LeNet模型
def create_lenet_model(input_shape):
    model = Sequential([
        Conv2D(6, (5, 5), activation='relu', input_shape=input_shape, padding='same'),
        MaxPooling2D((2, 2), strides=2),
        Conv2D(16, (5, 5), activation='relu'),
        MaxPooling2D((2, 2), strides=2),
        Flatten(),
        Dense(120, activation='relu'),
        Dense(84, activation='relu'),
        Dense(4, activation='softmax')
    ])
    model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
    return model

# 数据增强和数据生成器
train_datagen = ImageDataGenerator(rescale=1./255, validation_split=0.2)

# 训练数据生成器
train_generator = train_datagen.flow_from_directory(
    'D:\Pycharm_workspace\LeNet实验_四分类\image',
    target_size=(176, 208),
    batch_size=32,
    class_mode='categorical',
    subset='training'
)

# 验证数据生成器
validation_generator = train_datagen.flow_from_directory(
    'D:\Pycharm_workspace\LeNet实验_四分类\image',
    target_size=(176, 208),
    batch_size=32,
    class_mode='categorical',
    subset='validation'
)

# 创建并训练模型
input_shape = (176, 208, 3)
model = create_lenet_model(input_shape)
history = model.fit(train_generator, epochs=10, validation_data=validation_generator)

# 保存模型
model.save('dementia_classification_model.h5')

# 绘制训练和验证损失
plt.plot(history.history['loss'], label='训练损失')
plt.plot(history.history['val_loss'], label='验证损失')
plt.title('训练和验证损失')
plt.xlabel('时期')
plt.ylabel('损失')
plt.legend()
plt.show()

# 绘制训练和验证准确率
plt.plot(history.history['accuracy'], label='训练准确率')
plt.plot(history.history['val_accuracy'], label='验证准确率')
plt.title('训练和验证准确率')
plt.xlabel('时期')
plt.ylabel('准确率')
plt.legend()
plt.show()

loss值与准确率的变化图

可以看到才第四轮准确率就已经很高了

9 预测结果代码

import tensorflow as tf
from tensorflow.keras.models import load_model
from tensorflow.keras.preprocessing import image
import numpy as np
import matplotlib.pyplot as plt

# 加载模型
model = load_model('dementia_classification_model.h5')

# 定义类别标签
class_labels = ['MildDemented', 'ModerateDemented', 'NonDemented', 'VeryMildDemented']

# 预测函数
def predict_image(img_path):
    img = image.load_img(img_path, target_size=(176, 208))
    img_array = image.img_to_array(img)
    img_array = np.expand_dims(img_array, axis=0)
    img_array /= 255.0

    prediction = model.predict(img_array)
    predicted_class = class_labels[np.argmax(prediction)]

    # 显示图像和预测结果
    plt.imshow(image.load_img(img_path))
    plt.title(f'Predicted: {predicted_class}')
    plt.axis('off')
    plt.show()

# 预测并展示结果
img_path = r'D:\Pycharm_workspace\LeNet实验_四分类\image\VeryMildDemented\verymildDem0.jpg'  # 你的图片路径
predict_image(img_path)