Python机器学习设计——果蔬识别

一、选题背景

在学习了深度学习案例——MNIST手写数字识别和基于keras框架的猫狗图像识别，为了进一步熟悉tensorflow和keras的基本用法和网络框架，试想着实现对水果和蔬菜的种类进行识别。

二、机器学习案例设计方案

1.数据集来源

水果蔬菜数据集:

数据集来源于kaggle，因数据集作者为了构建一个应用程序，该应用程序可以从捕获的照片中识别食品，并为用户提供可以使用食品制作的不同食谱从必应图像搜索中抓取的，共包含4291张图像，36个种类，分为train(每个种类100张)，test(每个种类10张)，validation(每个种类10张)三个文件夹，上述每个文件夹都包含不同水果和蔬菜的子文件夹，其中存在相应食品的图像，包含水果10种，蔬菜26种，共36种。

2.采用的机器学习框架描述

卷积神经网络(Convolutional Neural Network, CNN)：卷积神经网络是一类包含卷积计算且具有深度结构的前馈神经网络，是深度学习技术中极具代表的网络结构之一，在图像处理领域取得了很大的成功，在国际标准的 ImageNet 数据集上，许多成功的模型都是基于 CNN 的。

Tensorflow：TensorFlow是谷歌基于DistBelief进行研发的第二代人工智能学习系统，其命名来源于本身的运行原理。Tensor（张量）意味着N维数组，Flow（流）意味着基于数据流图的计算，TensorFlow为张量从流图的一端流动到另一端计算过程。TensorFlow是将复杂的数据结构传输至人工智能神经网中进行分析和处理过程的系统，可被用于语音识别或图像识别等多项机器学习和深度学习领域。

Keras：Keras是一个模型级( model-level)的库，为开发深度学习模型提供了高层次的构建模块。它不处理张量操作、求微积分等基础的运算，而是依赖--个专门的、高度优化的张量库来完成这些运算。这个张量库就是Keras的后端引擎(backend engine)，如TensorFlow等。

3.涉及到的技术难点和解决思路

一些第三方库没有此模块的问题  解决思路：降版本或者使用功能相同的新模块

三、机器学习的实现步骤

果蔬识别

1.对原有数据集进行分类

1）将数据集train，test，和validation分类成fruit和vegetable两类

import os

import shutil

#训练集数据处理
def train(file_path):
    d=[]
    s=[]
    for root, dirs , files in os.walk(file_path): #读取文件并提取出文件路径和类名    
        d.append(root)#文件路径
        s.append(dirs)#类名
        for i in s:
            if i!=s[0]:
                s.remove(i)
        b = [ i for item in s for i in item]        
    d.pop(0)
    fruit=[ 'banana', 'apple', 'pear', 'grapes', 'orange','kiwi','watermelon', 'pomegranate', 'pineapple', 'mango']
    vegetables=['cucumber', 'carrot', 'capsicum', 'onion', 'potato', 'lemon', 'tomato', 
                'raddish', 'beetroot', 'cabbage', 'lettuce', 'spinach', 'soy beans', 
                'cauliflower', 'bell pepper', 'chilli pepper', 'turnip', 'corn', 'sweetcorn',
                'sweetpotato', 'paprika', 'jalepeno', 'ginger', 'garlic', 'peas', 'eggplant']
    #用for循环对水果和蔬菜进行分类
    #训练集数据处理
    #水果
    for i in b:
        for j in fruit:
            if i==j:
                shutil.move(os.path.join("C:/Users/linyicheng/Desktop/train",i),
                            "C:/Users/linyicheng/Desktop/fruit/train")
    #蔬菜
    for i in b:
        for j in vegetables:
            if i==j:
                shutil.move(os.path.join("C:/Users/linyicheng/Desktop/train",i),
                            "C:/Users/linyicheng/Desktop/vegetable/train")

train("C:/Users/linyicheng/Desktop/train")

#验证集数据处理
def validation(file_path):
    d=[]
    s=[]
    for root, dirs , files in os.walk(file_path):     
        d.append(root)
        s.append(dirs)
        for i in s:
            if i!=s[0]:
                s.remove(i)
        b = [ i for item in s for i in item]        
    d.pop(0)
    fruit=[ 'banana', 'apple', 'pear', 'grapes', 'orange','kiwi','watermelon', 'pomegranate', 'pineapple', 'mango']
    vegetables=['cucumber', 'carrot', 'capsicum', 'onion', 'potato', 'lemon', 'tomato', 
                'raddish', 'beetroot', 'cabbage', 'lettuce', 'spinach', 'soy beans', 
                'cauliflower', 'bell pepper', 'chilli pepper', 'turnip', 'corn', 'sweetcorn',
                'sweetpotato', 'paprika', 'jalepeno', 'ginger', 'garlic', 'peas', 'eggplant']
    #训练集数据处理
    #水果
    for i in b:
        for j in fruit:
            if i==j:
                shutil.move(os.path.join("C:/Users/linyicheng/Desktop/validation",i),
                            "C:/Users/linyicheng/Desktop/fruit/validation")
    #蔬菜
    for i in b:
        for j in vegetables:
            if i==j:
                shutil.move(os.path.join("C:/Users/linyicheng/Desktop/validation",i),
                            "C:/Users/linyicheng/Desktop/vegetable/validation")

validation("C:/Users/linyicheng/Desktop/validation")

#测试集数据处理
def test(file_path):
    d=[]
    s=[]
    for root, dirs , files in os.walk(file_path):     
        d.append(root)
        s.append(dirs)
        for i in s:
            if i!=s[0]:
                s.remove(i)
        b = [ i for item in s for i in item]        
    d.pop(0)
    fruit=[ 'banana', 'apple', 'pear', 'grapes', 'orange','kiwi','watermelon', 'pomegranate', 'pineapple', 'mango']
    vegetables=['cucumber', 'carrot', 'capsicum', 'onion', 'potato', 'lemon', 'tomato', 
                'raddish', 'beetroot', 'cabbage', 'lettuce', 'spinach', 'soy beans', 
                'cauliflower', 'bell pepper', 'chilli pepper', 'turnip', 'corn', 'sweetcorn',
                'sweetpotato', 'paprika', 'jalepeno', 'ginger', 'garlic', 'peas', 'eggplant']
    #训练集数据处理
    #水果
    for i in b:
        for j in fruit:
            if i==j:
                shutil.move(os.path.join("C:/Users/linyicheng/Desktop/test",i),
                            "C:/Users/linyicheng/Desktop/fruit/test")
    #蔬菜
    for i in b:
        for j in vegetables:
            if i==j:
                shutil.move(os.path.join("C:/Users/linyicheng/Desktop/test",i),
                            "C:/Users/linyicheng/Desktop/vegetable/test")

test("C:/Users/linyicheng/Desktop/test")

2.对数据集进行预处理，模仿猫狗识别建立数据集目录

1）对fruit，vegetable里的train，test，validation进行重命名

#水果
outer_path ="C:/Users/linyicheng/Desktop/fruit/test"
folderlist = os.listdir(outer_path)          #列举文件夹
 
for folder in folderlist:     
    inner_path = os.path.join(outer_path, folder)
    total_num_folder = len(folderlist)       #文件夹的总数
       #打印文件夹的总数
    filelist = os.listdir(inner_path)        #列举图片
    i = 0
    for item in filelist:
        total_num_file = len(filelist)       #单个文件夹内图片的总数
        if item.endswith('.jpg'):
            src = os.path.join(os.path.abspath(inner_path), item)           #原图的地址
            dst = os.path.join(os.path.abspath(inner_path), str(folder) + '_' + str(i) + '.jpg')        #新图的地址（这里可以把str(folder) + '_' + str(i) + '.jpg'改成你想改的名称）
            try:
                os.rename(src, dst)        
                i += 1
            except:
                continue

outer_path ="C:/Users/linyicheng/Desktop/fruit/train"
folderlist = os.listdir(outer_path)          #列举文件夹
 
for folder in folderlist:     
    inner_path = os.path.join(outer_path, folder)
    total_num_folder = len(folderlist)       #文件夹的总数
       #打印文件夹的总数
 
    filelist = os.listdir(inner_path)        #列举图片
    i = 0
    for item in filelist:
        total_num_file = len(filelist)       #单个文件夹内图片的总数
        if item.endswith('.jpg'):
            src = os.path.join(os.path.abspath(inner_path), item)           #原图的地址
            dst = os.path.join(os.path.abspath(inner_path), str(folder) + '_' + str(i) + '.jpg')        #新图的地址（这里可以把str(folder) + '_' + str(i) + '.jpg'改成你想改的名称）
            try:
                os.rename(src, dst)
                
                i += 1
            except:
                continue

#验证集
outer_path ="C:/Users/linyicheng/Desktop/fruit/validation"
folderlist = os.listdir(outer_path)          #列举文件夹
 
for folder in folderlist:     
    inner_path = os.path.join(outer_path, folder)
    total_num_folder = len(folderlist)       #文件夹的总数
       #打印文件夹的总数
 
    filelist = os.listdir(inner_path)        #列举图片
    i = 0
    for item in filelist:
        total_num_file = len(filelist)       #单个文件夹内图片的总数
        if item.endswith('.jpg'):
            src = os.path.join(os.path.abspath(inner_path), item)           #原图的地址
            dst = os.path.join(os.path.abspath(inner_path), str(folder) + '_' + str(i) + '.jpg')        #新图的地址（这里可以把str(folder) + '_' + str(i) + '.jpg'改成你想改的名称）
            try:
                os.rename(src, dst)
                
                i += 1
            except:
                continue

#蔬菜
outer_path ="C:/Users/linyicheng/Desktop/vegetable/test"
folderlist = os.listdir(outer_path)          #列举文件夹
 
for folder in folderlist:     
    inner_path = os.path.join(outer_path, folder)
    total_num_folder = len(folderlist)       #文件夹的总数
       #打印文件夹的总数
    filelist = os.listdir(inner_path)        #列举图片
    i = 0
    for item in filelist:
        total_num_file = len(filelist)       #单个文件夹内图片的总数
        if item.endswith('.jpg'):
            src = os.path.join(os.path.abspath(inner_path), item)           #原图的地址
            dst = os.path.join(os.path.abspath(inner_path), str(folder) + '_' + str(i) + '.jpg')        #新图的地址（这里可以把str(folder) + '_' + str(i) + '.jpg'改成你想改的名称）
            try:
                os.rename(src, dst)        
                i += 1
            except:
                continue

outer_path ="C:/Users/linyicheng/Desktop/vegetable/train"
folderlist = os.listdir(outer_path)          #列举文件夹
 
for folder in folderlist:     
    inner_path = os.path.join(outer_path, folder)
    total_num_folder = len(folderlist)       #文件夹的总数
       #打印文件夹的总数
 
    filelist = os.listdir(inner_path)        #列举图片
    i = 0
    for item in filelist:
        total_num_file = len(filelist)       #单个文件夹内图片的总数
        if item.endswith('.jpg'):
            src = os.path.join(os.path.abspath(inner_path), item)           #原图的地址
            dst = os.path.join(os.path.abspath(inner_path), str(folder) + '_' + str(i) + '.jpg')        #新图的地址（这里可以把str(folder) + '_' + str(i) + '.jpg'改成你想改的名称）
            try:
                os.rename(src, dst)
                
                i += 1
            except:
                continue

outer_path ="C:/Users/linyicheng/Desktop/vegetable/validation"
folderlist = os.listdir(outer_path)          #列举文件夹
 
for folder in folderlist:     
    inner_path = os.path.join(outer_path, folder)
    total_num_folder = len(folderlist)       #文件夹的总数
       #打印文件夹的总数
 
    filelist = os.listdir(inner_path)        #列举图片
    i = 0
    for item in filelist:
        total_num_file = len(filelist)       #单个文件夹内图片的总数
        if item.endswith('.jpg'):
            src = os.path.join(os.path.abspath(inner_path), item)           #原图的地址
            dst = os.path.join(os.path.abspath(inner_path), str(folder) + '_' + str(i) + '.jpg')        #新图的地址（这里可以把str(folder) + '_' + str(i) + '.jpg'改成你想改的名称）
            try:
                os.rename(src, dst)
                
                i += 1
            except:
                continue

2）将文件移到base目录中，构建数据集目录结构

#目标文件夹，此处为相对路径，也可以改为绝对路径
import shutil
determination = "C:/Users/linyicheng/Desktop/base/test/fruit"#目标
if not os.path.exists(determination):
    os.makedirs(determination)

#源文件夹路径
path = "C:/Users/linyicheng/Desktop/fruit/test"
folders = os.listdir(path)
for folder in folders:
    dir = path + '/' + str(folder)
    files = os.listdir(dir)
    for file in files:
        source = dir + '/' + str(file)
        deter = determination + '/' + str(file)
        shutil.copyfile(source, deter)

determination = "C:/Users/linyicheng/Desktop/base/train/fruit"
if not os.path.exists(determination):
    os.makedirs(determination)
path = "C:/Users/linyicheng/Desktop/fruit/train"
folders = os.listdir(path)
for folder in folders:
    dir = path + '/' + str(folder)
    files = os.listdir(dir)
    for file in files:
        source = dir + '/' + str(file)
        deter = determination + '/' + str(file)
        shutil.copyfile(source, deter)

determination = "C:/Users/linyicheng/Desktop/base/validation/fruit"
if not os.path.exists(determination):
    os.makedirs(determination)
path = "C:/Users/linyicheng/Desktop/fruit/validation"
folders = os.listdir(path)
for folder in folders:
    dir = path + '/' + str(folder)
    files = os.listdir(dir)
    for file in files:
        source = dir + '/' + str(file)
        deter = determination + '/' + str(file)
        shutil.copyfile(source, deter)

import shutil
determination = "C:/Users/linyicheng/Desktop/base/test/vegetable"#目标
if not os.path.exists(determination):
    os.makedirs(determination)

#源文件夹路径
path = "C:/Users/linyicheng/Desktop/vegetable/test"
folders = os.listdir(path)
for folder in folders:
    dir = path + '/' + str(folder)
    files = os.listdir(dir)
    for file in files:
        source = dir + '/' + str(file)
        deter = determination + '/' + str(file)
        shutil.copyfile(source, deter)

determination = "C:/Users/linyicheng/Desktop/base/train/vegetable"
if not os.path.exists(determination):
    os.makedirs(determination)
path = "C:/Users/linyicheng/Desktop/vegetable/train"
folders = os.listdir(path)
for folder in folders:
    dir = path + '/' + str(folder)
    files = os.listdir(dir)
    for file in files:
        source = dir + '/' + str(file)
        deter = determination + '/' + str(file)
        shutil.copyfile(source, deter)

determination = "C:/Users/linyicheng/Desktop/base/validation/vegetable"
if not os.path.exists(determination):
    os.makedirs(determination)
path = "C:/Users/linyicheng/Desktop/vegetable/validation"
folders = os.listdir(path)
for folder in folders:
    dir = path + '/' + str(folder)
    files = os.listdir(dir)
    for file in files:
        source = dir + '/' + str(file)
        deter = determination + '/' + str(file)
        shutil.copyfile(source, deter)

3.设置路径

base_dir="C:/Users/linyicheng/Desktop/base/"
train_data_dir="C:/Users/linyicheng/Desktop/base/train/"
test_data_dir="C:/Users/linyicheng/Desktop/base/test/"
val_data_dir="C:/Users/linyicheng/Desktop/base/validation/"
train_fruit_data="C:/Users/linyicheng/Desktop/base/train/fruit/"
test_fruit_data="C:/Users/linyicheng/Desktop/base/test/fruit/"
test_vegetable_data="C:/Users/linyicheng/Desktop/base/test/vegetable/"

4.搭建卷积神经网络

#搭建卷积神经网络
from keras import layers
from keras import models
model=models.Sequential()
"""
Output shape计算公式：(输入尺寸-卷积核尺寸/步长+1
对CNN模型，Param的计算方法如下：
卷积核长度*卷积核宽度*通道数+1）*卷积核个数
输出图片尺寸：150-3+1=148*148
参数数量：32*3*3*3+32=896
"""
model.add(layers.Conv2D(32,(3,3),activation='relu',input_shape=(150,150,3)))#卷积层1
model.add(layers.MaxPooling2D(2,2))#最大池化层1
# 输出图片尺寸：148/2=74*74 
# 输出图片尺寸：74-3+1=72*72，参数数量：64*3*3*32+64=18496
#32是第1个卷积层的输出的通道数
model.add(layers.Conv2D(64, (3, 3), activation='relu'))#卷积层2
model.add(layers.MaxPooling2D((2, 2)))#最大池化层2
# 输出图片尺寸：72/2=36*36
 
#Output Shape的输出为36
# 输出图片尺寸：36-3+1=34*34，参数数量：128*3*3*64+128=73856
model.add(layers.Conv2D(128, (3, 3), activation='relu'))#卷积层3
model.add(layers.MaxPooling2D((2, 2)))#最大池化层3
# 输出图片尺寸：34/2=17*17
# 输出图片尺寸：17-3+1=15*15，参数数量：128*3*3*128+128=147584
model.add(layers.Conv2D(128, (3, 3), activation='relu'))# 输出图片尺寸：15/2=7*7
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Flatten())
model.add(layers.Dense(512, activation='relu'))   #全连接层1
model.add(layers.Dense(1, activation='sigmoid'))#全连接层2，作为输出层

 从输出看出，卷积神经网络dense_1 (Dense)的参数总数达到300多万

model.summary() 

 5.编译模型

# 编译模型
# RMSprop 优化器。因为网络最后一层是单一sigmoid单元
from keras import optimizers
model.compile(loss='binary_crossentropy',
              optimizer=optimizers.RMSprop(lr=1e-4),
               metrics=['acc'])

6.使用 ImageDataGenerator 从目录中读取样本数据

from keras.preprocessing.image import ImageDataGenerator
#归一化

train_datagen = ImageDataGenerator(rescale=1./255)
validation_datagen=ImageDataGenerator(rescale=1./255)
test_datagen = ImageDataGenerator(rescale=1./255)
# 数据集加载函数，指明数据集的位置并统一处理为imgheight*imgwidth的大小，同时设置batch
    # 加载训练集
train_ds = train_datagen.flow_from_directory (
        #文件目录位置
        train_data_dir,
#输入训练图像尺寸,所有图片的size必须是150x150
        class_mode='binary',
        target_size=(150, 150),
        batch_size=20)
    #加载测试集
test_ds = test_datagen.flow_from_directory(
        test_data_dir,
        class_mode='binary',
        target_size=(150, 150),
        batch_size=20)
    # 加载验证集
val_ds =test_datagen.flow_from_directory(
        val_data_dir,
        class_mode='binary',
        target_size=(150, 150),
        batch_size=20)

for data_batch, labels_batch in train_ds:
    print('data batch shape:', data_batch.shape)
    print('labels batch shape:', labels_batch.shape)
    break

 

 7.训练模型（30次）

#训练模型30轮次，可以修改epochs的值
history = model.fit_generator(
       train_ds,
       steps_per_epoch=100,
       epochs=30,
       validation_data=val_ds,
       validation_steps=50)

可以看出训练集精度（acc）：0.9130 和验证集精度（val_acc）：0.9803

模型较为成功，接下来通过数据增强来提升精度

 #将训练过程产生的数据保存为h5文件

model.save('fruit_and_vegetable_30epoch.h5') 

8.绘制训练过程中的损失曲线和精度曲线

import matplotlib.pyplot as plt
 
acc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']

epochs = range(1, len(acc) + 1)

plt.plot(epochs, acc, 'bo', label='Training acc')
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and validation accuracy')
plt.legend()
plt.figure() 
plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss')
plt.legend()
 
plt.show()

由图可知训练精度随着时间线性增加，直到接近100%

9.单张图片进行判断图片是水果还是蔬菜

from PIL import Image
from keras.preprocessing import image
from keras.models import load_model
import numpy as np
#加载模型
model=load_model('fruit_and_vegetable_30epoch.h5')
#本地图片路径
img_path=("C:/Users/linyicheng/Desktop/base/train/fruit/apple_41.jpg")
img = image.load_img(img_path,target_size=(150,150))
# 将其转换为具有形状的Numpy数组（150、150、3）
img_tensor = image.img_to_array(img)/255
# 将其形态变为(150,150,150,3)的形状
img_tensor = np.expand_dims(img_tensor, axis=0)
#取图片信息
prediction =model.predict(img_tensor)
#输出识别率
print(prediction)
if prediction>0.5:
    print('水果')
else:
    print('蔬菜')
plt.imshow(img)

10. 数据增强训练，利用ImageDataGenerator实现数据增强

from keras.preprocessing.image import ImageDataGenerator
from keras.models import load_model
from tensorflow.python.keras.preprocessing import image
from tensorflow.python.keras.preprocessing.image import ImageDataGenerator,array_to_img,img_to_array,load_img

import os
from keras.preprocessing.image import load_img
import matplotlib.pyplot as plt
figure,ax = plt.subplots(nrows=1, ncols=4, sharex=True, 
                         sharey=True,figsize=(16,10))

datagen=ImageDataGenerator(
    rotation_range=40,      #图像随机旋转角度范围
    width_shift_range=0.2,  #图片在水平方向上平移的比例
    height_shift_range=0.2, #图片在垂直方向上平移的比例
    shear_range=0.2,        #随机错切变换的角度
    zoom_range=0.2,         #图像随机缩放的范围
    horizontal_flip=True,   #随机将一半图像水平翻转
      
    fill_mode='nearest')
#填充创建像素的一种方法
fnames=[os.path.join(train_fruit_data,fname)for fname in os.listdir(train_fruit_data)]
img_path=fnames[6]
img=image.load_img(img_path,target_size=(150,150))
x = image.img_to_array(img)/255
x = x.reshape((1,) + x.shape)
i = 0
for batch in datagen.flow(x, batch_size=1):
    ax = ax.flatten()   # 将子图从多维变成一维
    plt.figure(i)
    imgplot = ax[i-1].imshow(image.array_to_img(batch[0]), cmap='Greys', interpolation='nearest')
    i += 1
    if i % 4 == 0:
        break
plt.show() #绘制增强后的图像
'''
虽然使用了数据增强技术，但是从输出的图像来看，图片之间还是有很大的相似度，
因为它们均来自同一张原始图片，并没有提供新的信息。
为了尽可能消除过拟合，可以在模型中增加一个DDropout层，添加到密集连接分类
器前
'''

11. 在紧密连接的分类器之前为模型添加一个Dropout层

#在紧密连接的分类器之前为模型添加一个Dropout层
model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='relu',
                        input_shape=(150, 150, 3)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Flatten())
model.add(layers.Dropout(0.5))
model.add(layers.Dense(512, activation='relu'))
model.add(layers.Dense(1, activation='sigmoid'))

12.编译模型

from keras import optimizers
model.compile(loss='binary_crossentropy',
              optimizer=optimizers.RMSprop(lr=1e-4),
               metrics=['acc'])

13.使用数据增强和dropout来训练我们的网络

#使用数据增强和dropout来训练我们的网络
#归一化
train_datagen = ImageDataGenerator(
     rescale=1./255,
     rotation_range=40,
     width_shift_range=0.2,
     height_shift_range=0.2,
     shear_range=0.2,
     zoom_range=0.2,
     horizontal_flip=True)
test_datagen = ImageDataGenerator(rescale=1./255)
train_generator = train_datagen.flow_from_directory(
     train_data_dir,
     #输入训练图像尺寸,所有图片的size必须是150x150
     target_size=(150, 150),
     batch_size=32,
     #因为我们使用二元交叉熵损失，我们需要二元标签
     class_mode='binary')
validation_generator = test_datagen.flow_from_directory(
     val_data_dir,
     target_size=(150, 150),
     batch_size=32,
     class_mode='binary')

 14.训练模型（100次）

history = model.fit_generator(
       train_generator,
       steps_per_epoch=100,
       epochs=100,
       validation_data=validation_generator,
       validation_steps=50)

 model.save('fruit_and_vegetable_100epoch.h5')  

15.绘制训练过程中的损失曲线和精度曲线

import matplotlib.pyplot as plt
 
acc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']

epochs = range(1, len(acc) + 1)

plt.plot(epochs, acc, 'bo', label='Training acc')
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and validation accuracy')
plt.legend()
plt.figure() 
plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss')
plt.legend()
 
plt.show()

 16.使用新的训练模型进行识别

from keras.preprocessing import image

from keras.models import load_model
import numpy as np
#加载模型
model=load_model('fruit_and_vegetable_100epoch.h5')
#本地图片路径
img_path=("C:/Users/linyicheng/Desktop/base/train/fruit/apple_42.jpg")
img = image.load_img(img_path,target_size=(150,150))
# 将其转换为具有形状的Numpy数组（150、150、3）
img_tensor = image.img_to_array(img)/255
# 将其形态变为(150,150,150,3)的形状
img_tensor = np.expand_dims(img_tensor, axis=0)
#取图片信息
prediction =model.predict(img_tensor)
#输出识别率
print(prediction)
if prediction>0.5:
    print('水果')
else:
    print('蔬菜')
plt.imshow(img)

17.多张图片进行识别判断

 

from PIL import Image
from keras.preprocessing import image
import matplotlib.pyplot as plt
from keras.models import load_model
plt.rcParams['font.family'] = ['SimHei']
figure,ax = plt.subplots(nrows=2, ncols=4, sharex=True, sharey=True,figsize=(24,10))
def convertjpg(jpgfile):
    #将图片缩小到（150，150）的大小
    img=Image.open(jpgfile)
    try:
        new_img=img.resize((150,150),Image.BILINEAR)
        return new_img
    except Exception as e:
        print(e)
snames=[os.path.join(test_vegetable_data,sname)for sname in os.listdir(test_vegetable_data)]
j=0
for i in snames[0:8] :
    j+=1
    img_show=convertjpg(i)
    img_scale = image.img_to_array(img_show)
    img_scale = img_scale.reshape(1,150,150,3) # 将形状转化为(1,150,150,3)
    img_scale = img_scale.astype('float32')/255
    result = model.predict(img_scale)# 预测函数-
    if result>0.5:
        title = '蔬菜'
        label = '正确识别率为：'+str(result)
    else:
        title = '水果'
        label = '正确识别率为：'+str(1-result)
         
    ax = ax.flatten()   # 将子图从多维变成一维
    ax[j-1].imshow(img_show, cmap='Greys', interpolation='nearest')
     # 子图标题
    ax[j-1].set_title(title,fontsize=24)
     # 子图X轴标签
    ax[j-1].set_xlabel(label,fontsize=24)
     # 去掉刻度
    ax[0].set_xticks([])
    ax[0].set_yticks([]) 
plt.show()

figure,ax = plt.subplots(nrows=2, ncols=4, sharex=True, sharey=True,figsize=(24,10))

fnames=[os.path.join(test_fruit_data,fname)for fname in os.listdir(test_fruit_data)]
j=0
for i in fnames[0:8]: 
    j+=1
    img_show=convertjpg(i)
    img_scale = image.img_to_array(img_show)
    img_scale = img_scale.reshape(1,150,150,3) # 将形状转化为(1,150,150,3)
    img_scale = img_scale.astype('float32')/255
    result = model.predict(img_scale)# 预测函数-
    if result>0.5:
        title = '蔬菜'
        label = '正确识别率为：'+ str(result)
    else:
        title = '水果'
        label = '正确识别率为：'+str(1-result)         
    ax = ax.flatten()   # 将子图从多维变成一维
    ax[j-1].imshow(img_show, cmap='Greys', interpolation='nearest')
     # 子图标题
    ax[j-1].set_title(title,fontsize=24)
     # 子图X轴标签
    ax[j-1].set_xlabel(label,fontsize=24)
    # 去掉刻度
    ax[0].set_xticks([])
    ax[0].set_yticks([])

 全代码附上

import os

import shutil

#训练集数据处理
def train(file_path):
    d=[]
    s=[]
    for root, dirs , files in os.walk(file_path): #读取文件并提取出文件路径和类名    
        d.append(root)#文件路径
        s.append(dirs)#类名
        for i in s:
            if i!=s[0]:
                s.remove(i)
        b = [ i for item in s for i in item]        
    d.pop(0)
    fruit=[ 'banana', 'apple', 'pear', 'grapes', 'orange','kiwi','watermelon', 'pomegranate', 'pineapple', 'mango']
    vegetables=['cucumber', 'carrot', 'capsicum', 'onion', 'potato', 'lemon', 'tomato', 
                'raddish', 'beetroot', 'cabbage', 'lettuce', 'spinach', 'soy beans', 
                'cauliflower', 'bell pepper', 'chilli pepper', 'turnip', 'corn', 'sweetcorn',
                'sweetpotato', 'paprika', 'jalepeno', 'ginger', 'garlic', 'peas', 'eggplant']
    #用for循环对水果和蔬菜进行分类
    #训练集数据处理
    #水果
    for i in b:
        for j in fruit:
            if i==j:
                shutil.move(os.path.join("C:/Users/linyicheng/Desktop/train",i),
                            "C:/Users/linyicheng/Desktop/fruit/train")
    #蔬菜
    for i in b:
        for j in vegetables:
            if i==j:
                shutil.move(os.path.join("C:/Users/linyicheng/Desktop/train",i),
                            "C:/Users/linyicheng/Desktop/vegetable/train")

train("C:/Users/linyicheng/Desktop/train")
#验证集数据处理
def validation(file_path):
    d=[]
    s=[]
    for root, dirs , files in os.walk(file_path):     
        d.append(root)
        s.append(dirs)
        for i in s:
            if i!=s[0]:
                s.remove(i)
        b = [ i for item in s for i in item]        
    d.pop(0)
    fruit=[ 'banana', 'apple', 'pear', 'grapes', 'orange','kiwi','watermelon', 'pomegranate', 'pineapple', 'mango']
    vegetables=['cucumber', 'carrot', 'capsicum', 'onion', 'potato', 'lemon', 'tomato', 
                'raddish', 'beetroot', 'cabbage', 'lettuce', 'spinach', 'soy beans', 
                'cauliflower', 'bell pepper', 'chilli pepper', 'turnip', 'corn', 'sweetcorn',
                'sweetpotato', 'paprika', 'jalepeno', 'ginger', 'garlic', 'peas', 'eggplant']
    #训练集数据处理
    #水果
    for i in b:
        for j in fruit:
            if i==j:
                shutil.move(os.path.join("C:/Users/linyicheng/Desktop/validation",i),
                            "C:/Users/linyicheng/Desktop/fruit/validation")
    #蔬菜
    for i in b:
        for j in vegetables:
            if i==j:
                shutil.move(os.path.join("C:/Users/linyicheng/Desktop/validation",i),
                            "C:/Users/linyicheng/Desktop/vegetable/validation")

validation("C:/Users/linyicheng/Desktop/validation")
#测试集数据处理
def test(file_path):
    d=[]
    s=[]
    for root, dirs , files in os.walk(file_path):     
        d.append(root)
        s.append(dirs)
        for i in s:
            if i!=s[0]:
                s.remove(i)
        b = [ i for item in s for i in item]        
    d.pop(0)
    fruit=[ 'banana', 'apple', 'pear', 'grapes', 'orange','kiwi','watermelon', 'pomegranate', 'pineapple', 'mango']
    vegetables=['cucumber', 'carrot', 'capsicum', 'onion', 'potato', 'lemon', 'tomato', 
                'raddish', 'beetroot', 'cabbage', 'lettuce', 'spinach', 'soy beans', 
                'cauliflower', 'bell pepper', 'chilli pepper', 'turnip', 'corn', 'sweetcorn',
                'sweetpotato', 'paprika', 'jalepeno', 'ginger', 'garlic', 'peas', 'eggplant']
    #训练集数据处理
    #水果
    for i in b:
        for j in fruit:
            if i==j:
                shutil.move(os.path.join("C:/Users/linyicheng/Desktop/test",i),
                            "C:/Users/linyicheng/Desktop/fruit/test")
    #蔬菜
    for i in b:
        for j in vegetables:
            if i==j:
                shutil.move(os.path.join("C:/Users/linyicheng/Desktop/test",i),
                            "C:/Users/linyicheng/Desktop/vegetable/test")

test("C:/Users/linyicheng/Desktop/test")
#水果
outer_path ="C:/Users/linyicheng/Desktop/fruit/test"
folderlist = os.listdir(outer_path)          #列举文件夹
 
for folder in folderlist:     
    inner_path = os.path.join(outer_path, folder)
    total_num_folder = len(folderlist)       #文件夹的总数
       #打印文件夹的总数
    filelist = os.listdir(inner_path)        #列举图片
    i = 0
    for item in filelist:
        total_num_file = len(filelist)       #单个文件夹内图片的总数
        if item.endswith('.jpg'):
            src = os.path.join(os.path.abspath(inner_path), item)           #原图的地址
            dst = os.path.join(os.path.abspath(inner_path), str(folder) + '_' + str(i) + '.jpg')        #新图的地址（这里可以把str(folder) + '_' + str(i) + '.jpg'改成你想改的名称）
            try:
                os.rename(src, dst)        
                i += 1
            except:
                continue
outer_path ="C:/Users/linyicheng/Desktop/fruit/train"
folderlist = os.listdir(outer_path)          #列举文件夹
 
for folder in folderlist:     
    inner_path = os.path.join(outer_path, folder)
    total_num_folder = len(folderlist)       #文件夹的总数
       #打印文件夹的总数
 
    filelist = os.listdir(inner_path)        #列举图片
    i = 0
    for item in filelist:
        total_num_file = len(filelist)       #单个文件夹内图片的总数
        if item.endswith('.jpg'):
            src = os.path.join(os.path.abspath(inner_path), item)           #原图的地址
            dst = os.path.join(os.path.abspath(inner_path), str(folder) + '_' + str(i) + '.jpg')        #新图的地址（这里可以把str(folder) + '_' + str(i) + '.jpg'改成你想改的名称）
            try:
                os.rename(src, dst)
                
                i += 1
            except:
                continue
outer_path ="C:/Users/linyicheng/Desktop/fruit/validation"
folderlist = os.listdir(outer_path)          #列举文件夹
 
for folder in folderlist:     
    inner_path = os.path.join(outer_path, folder)
    total_num_folder = len(folderlist)       #文件夹的总数
       #打印文件夹的总数
 
    filelist = os.listdir(inner_path)        #列举图片
    i = 0
    for item in filelist:
        total_num_file = len(filelist)       #单个文件夹内图片的总数
        if item.endswith('.jpg'):
            src = os.path.join(os.path.abspath(inner_path), item)           #原图的地址
            dst = os.path.join(os.path.abspath(inner_path), str(folder) + '_' + str(i) + '.jpg')        #新图的地址（这里可以把str(folder) + '_' + str(i) + '.jpg'改成你想改的名称）
            try:
                os.rename(src, dst)
                
                i += 1
            except:
                continue
#蔬菜
outer_path ="C:/Users/linyicheng/Desktop/vegetable/test"
folderlist = os.listdir(outer_path)          #列举文件夹
 
for folder in folderlist:     
    inner_path = os.path.join(outer_path, folder)
    total_num_folder = len(folderlist)       #文件夹的总数
       #打印文件夹的总数
    filelist = os.listdir(inner_path)        #列举图片
    i = 0
    for item in filelist:
        total_num_file = len(filelist)       #单个文件夹内图片的总数
        if item.endswith('.jpg'):
            src = os.path.join(os.path.abspath(inner_path), item)           #原图的地址
            dst = os.path.join(os.path.abspath(inner_path), str(folder) + '_' + str(i) + '.jpg')        #新图的地址（这里可以把str(folder) + '_' + str(i) + '.jpg'改成你想改的名称）
            try:
                os.rename(src, dst)        
                i += 1
            except:
                continue
outer_path ="C:/Users/linyicheng/Desktop/vegetable/train"
folderlist = os.listdir(outer_path)          #列举文件夹
 
for folder in folderlist:     
    inner_path = os.path.join(outer_path, folder)
    total_num_folder = len(folderlist)       #文件夹的总数
       #打印文件夹的总数
 
    filelist = os.listdir(inner_path)        #列举图片
    i = 0
    for item in filelist:
        total_num_file = len(filelist)       #单个文件夹内图片的总数
        if item.endswith('.jpg'):
            src = os.path.join(os.path.abspath(inner_path), item)           #原图的地址
            dst = os.path.join(os.path.abspath(inner_path), str(folder) + '_' + str(i) + '.jpg')        #新图的地址（这里可以把str(folder) + '_' + str(i) + '.jpg'改成你想改的名称）
            try:
                os.rename(src, dst)
                
                i += 1
            except:
                continue
outer_path ="C:/Users/linyicheng/Desktop/vegetable/validation"
folderlist = os.listdir(outer_path)          #列举文件夹
 
for folder in folderlist:     
    inner_path = os.path.join(outer_path, folder)
    total_num_folder = len(folderlist)       #文件夹的总数
       #打印文件夹的总数
 
    filelist = os.listdir(inner_path)        #列举图片
    i = 0
    for item in filelist:
        total_num_file = len(filelist)       #单个文件夹内图片的总数
        if item.endswith('.jpg'):
            src = os.path.join(os.path.abspath(inner_path), item)           #原图的地址
            dst = os.path.join(os.path.abspath(inner_path), str(folder) + '_' + str(i) + '.jpg')        #新图的地址（这里可以把str(folder) + '_' + str(i) + '.jpg'改成你想改的名称）
            try:
                os.rename(src, dst)
                
                i += 1
            except:
                continue
#目标文件夹，此处为相对路径，也可以改为绝对路径
import shutil
determination = "C:/Users/linyicheng/Desktop/base/test/fruit"#目标
if not os.path.exists(determination):
    os.makedirs(determination)

#源文件夹路径
path = "C:/Users/linyicheng/Desktop/fruit/test"
folders = os.listdir(path)
for folder in folders:
    dir = path + '/' + str(folder)
    files = os.listdir(dir)
    for file in files:
        source = dir + '/' + str(file)
        deter = determination + '/' + str(file)
        shutil.copyfile(source, deter)
determination = "C:/Users/linyicheng/Desktop/base/train/fruit"
if not os.path.exists(determination):
    os.makedirs(determination)
path = "C:/Users/linyicheng/Desktop/fruit/train"
folders = os.listdir(path)
for folder in folders:
    dir = path + '/' + str(folder)
    files = os.listdir(dir)
    for file in files:
        source = dir + '/' + str(file)
        deter = determination + '/' + str(file)
        shutil.copyfile(source, deter)
determination = "C:/Users/linyicheng/Desktop/base/validation/fruit"
if not os.path.exists(determination):
    os.makedirs(determination)
path = "C:/Users/linyicheng/Desktop/fruit/validation"
folders = os.listdir(path)
for folder in folders:
    dir = path + '/' + str(folder)
    files = os.listdir(dir)
    for file in files:
        source = dir + '/' + str(file)
        deter = determination + '/' + str(file)
        shutil.copyfile(source, deter)
import shutil
determination = "C:/Users/linyicheng/Desktop/base/test/vegetable"#目标
if not os.path.exists(determination):
    os.makedirs(determination)

#源文件夹路径
path = "C:/Users/linyicheng/Desktop/vegetable/test"
folders = os.listdir(path)
for folder in folders:
    dir = path + '/' + str(folder)
    files = os.listdir(dir)
    for file in files:
        source = dir + '/' + str(file)
        deter = determination + '/' + str(file)
        shutil.copyfile(source, deter)
determination = "C:/Users/linyicheng/Desktop/base/train/vegetable"
if not os.path.exists(determination):
    os.makedirs(determination)
path = "C:/Users/linyicheng/Desktop/vegetable/train"
folders = os.listdir(path)
for folder in folders:
    dir = path + '/' + str(folder)
    files = os.listdir(dir)
    for file in files:
        source = dir + '/' + str(file)
        deter = determination + '/' + str(file)
        shutil.copyfile(source, deter)
determination = "C:/Users/linyicheng/Desktop/base/validation/vegetable"
if not os.path.exists(determination):
    os.makedirs(determination)
path = "C:/Users/linyicheng/Desktop/vegetable/validation"
folders = os.listdir(path)
for folder in folders:
    dir = path + '/' + str(folder)
    files = os.listdir(dir)
    for file in files:
        source = dir + '/' + str(file)
        deter = determination + '/' + str(file)
        shutil.copyfile(source, deter)

base_dir="C:/Users/linyicheng/Desktop/base/"
train_data_dir="C:/Users/linyicheng/Desktop/base/train/"
test_data_dir="C:/Users/linyicheng/Desktop/base/test/"
val_data_dir="C:/Users/linyicheng/Desktop/base/validation/"
train_fruit_data="C:/Users/linyicheng/Desktop/base/train/fruit/"
test_fruit_data="C:/Users/linyicheng/Desktop/base/test/fruit/"
test_vegetable_data="C:/Users/linyicheng/Desktop/base/test/vegetable/"

#搭建卷积神经网络
from keras import layers
from keras import models
model=models.Sequential()
"""
Output shape计算公式：(输入尺寸-卷积核尺寸/步长+1
对CNN模型，Param的计算方法如下：
卷积核长度*卷积核宽度*通道数+1）*卷积核个数
输出图片尺寸：150-3+1=148*148
参数数量：32*3*3*3+32=896
"""
model.add(layers.Conv2D(32,(3,3),activation='relu',input_shape=(150,150,3)))#卷积层1
model.add(layers.MaxPooling2D(2,2))#最大池化层1
# 输出图片尺寸：148/2=74*74 
# 输出图片尺寸：74-3+1=72*72，参数数量：64*3*3*32+64=18496
#32是第1个卷积层的输出的通道数
model.add(layers.Conv2D(64, (3, 3), activation='relu'))#卷积层2
model.add(layers.MaxPooling2D((2, 2)))#最大池化层2
# 输出图片尺寸：72/2=36*36
 
#Output Shape的输出为36
# 输出图片尺寸：36-3+1=34*34，参数数量：128*3*3*64+128=73856
model.add(layers.Conv2D(128, (3, 3), activation='relu'))#卷积层3
model.add(layers.MaxPooling2D((2, 2)))#最大池化层3
# 输出图片尺寸：34/2=17*17
# 输出图片尺寸：17-3+1=15*15，参数数量：128*3*3*128+128=147584
model.add(layers.Conv2D(128, (3, 3), activation='relu'))# 输出图片尺寸：15/2=7*7
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Flatten())
model.add(layers.Dense(512, activation='relu'))   #全连接层1
model.add(layers.Dense(1, activation='sigmoid'))#全连接层2，作为输出层
model.summary()

from keras import optimizers
model.compile(loss='binary_crossentropy',
              optimizer=optimizers.RMSprop(lr=1e-4),
               metrics=['acc'])

from keras.preprocessing.image import ImageDataGenerator
#归一化

train_datagen = ImageDataGenerator(rescale=1./255)
validation_datagen=ImageDataGenerator(rescale=1./255)
test_datagen = ImageDataGenerator(rescale=1./255)
# 数据集加载函数，指明数据集的位置并统一处理为imgheight*imgwidth的大小，同时设置batch
    # 加载训练集
train_ds = train_datagen.flow_from_directory (
        train_data_dir,
        class_mode='binary',
        target_size=(150, 150),
        batch_size=20)
    #加载测试集
test_ds = test_datagen.flow_from_directory(
        test_data_dir,
        class_mode='binary',
        target_size=(150, 150),
        batch_size=20)
    # 加载验证集
val_ds =test_datagen.flow_from_directory(
        val_data_dir,
        class_mode='binary',
        target_size=(150, 150),
        batch_size=20)

for data_batch, labels_batch in train_ds:
    print('data batch shape:', data_batch.shape)
    print('labels batch shape:', labels_batch.shape)
    break

#训练模型30轮次，可以修改epochs的值
history = model.fit_generator(
       train_ds,
       steps_per_epoch=100,
       epochs=30,
       validation_data=val_ds,
       validation_steps=50)

 #将训练过程产生的数据保存为h5文件
model.save('fruit_and_vegetable_30epoch.h5')

import matplotlib.pyplot as plt
 
acc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']

epochs = range(1, len(acc) + 1)

plt.plot(epochs, acc, 'bo', label='Training acc')
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and validation accuracy')
plt.legend()
plt.figure() 
plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss')
plt.legend()
 
plt.show()

from PIL import Image
from keras.preprocessing import image
from keras.models import load_model
import numpy as np
#加载模型
model=load_model('fruit_and_vegetable_30epoch.h5')
#本地图片路径
img_path=("C:/Users/linyicheng/Desktop/base/train/fruit/apple_41.jpg")
img = image.load_img(img_path,target_size=(150,150))
# 将其转换为具有形状的Numpy数组（150、150、3）
img_tensor = image.img_to_array(img)/255
# 将其形态变为(150,150,150,3)的形状
img_tensor = np.expand_dims(img_tensor, axis=0)
#取图片信息
prediction =model.predict(img_tensor)
#输出识别率
print(prediction)
if prediction>0.5:
    print('水果')
else:
    print('蔬菜')
plt.imshow(img)

from keras.preprocessing.image import ImageDataGenerator
from keras.models import load_model
from tensorflow.python.keras.preprocessing import image
from tensorflow.python.keras.preprocessing.image import ImageDataGenerator,array_to_img,img_to_array,load_img

import os
from keras.preprocessing.image import load_img
import matplotlib.pyplot as plt
figure,ax = plt.subplots(nrows=1, ncols=4, sharex=True, 
                         sharey=True,figsize=(16,10))

datagen=ImageDataGenerator(
    rotation_range=40,      #图像随机旋转角度范围
    width_shift_range=0.2,  #图片在水平方向上平移的比例
    height_shift_range=0.2, #图片在垂直方向上平移的比例
    shear_range=0.2,        #随机错切变换的角度
    zoom_range=0.2,         #图像随机缩放的范围
    horizontal_flip=True,   #随机将一半图像水平翻转
      
    fill_mode='nearest')
#填充创建像素的一种方法
fnames=[os.path.join(train_fruit_data,fname)for fname in os.listdir(train_fruit_data)]
img_path=fnames[6]
img=image.load_img(img_path,target_size=(150,150))
x = image.img_to_array(img)/255
x = x.reshape((1,) + x.shape)
i = 0
for batch in datagen.flow(x, batch_size=1):
    ax = ax.flatten()   # 将子图从多维变成一维
    plt.figure(i)
    imgplot = ax[i-1].imshow(image.array_to_img(batch[0]), cmap='Greys', interpolation='nearest')
    i += 1
    if i % 4 == 0:
        break
plt.show() #绘制增强后的图像
'''
虽然使用了数据增强技术，但是从输出的图像来看，图片之间还是有很大的相似度，
因为它们均来自同一张原始图片，并没有提供新的信息。
为了尽可能消除过拟合，可以在模型中增加一个DDropout层，添加到密集连接分类
器前
'''
#在紧密连接的分类器之前为模型添加一个Dropout层
model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='relu',
                        input_shape=(150, 150, 3)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Flatten())
model.add(layers.Dropout(0.5))
model.add(layers.Dense(512, activation='relu'))
model.add(layers.Dense(1, activation='sigmoid'))

from keras import optimizers
model.compile(loss='binary_crossentropy',
              optimizer=optimizers.RMSprop(lr=1e-4),
               metrics=['acc'])

#使用数据增强和dropout来训练我们的网络
#归一化
train_datagen = ImageDataGenerator(
     rescale=1./255,
     rotation_range=40,
     width_shift_range=0.2,
     height_shift_range=0.2,
     shear_range=0.2,
     zoom_range=0.2,
     horizontal_flip=True)
test_datagen = ImageDataGenerator(rescale=1./255)
train_generator = train_datagen.flow_from_directory(
     train_data_dir,
     #输入训练图像尺寸,所有图片的size必须是150x150
     target_size=(150, 150),
     batch_size=32,
     #因为我们使用二元交叉熵损失，我们需要二元标签
     class_mode='binary')
validation_generator = test_datagen.flow_from_directory(
     val_data_dir,
     target_size=(150, 150),
     batch_size=32,
     class_mode='binary')

history = model.fit_generator(
       train_generator,
       steps_per_epoch=100,
       epochs=100,
       validation_data=validation_generator,
       validation_steps=50)

model.save('fruit_and_vegetable_100epoch.h5')

import matplotlib.pyplot as plt
 
acc = history.history['acc']
val_acc = history.history['val_acc']
loss = history.history['loss']
val_loss = history.history['val_loss']

epochs = range(1, len(acc) + 1)

plt.plot(epochs, acc, 'bo', label='Training acc')
plt.plot(epochs, val_acc, 'b', label='Validation acc')
plt.title('Training and validation accuracy')
plt.legend()
plt.figure() 
plt.plot(epochs, loss, 'bo', label='Training loss')
plt.plot(epochs, val_loss, 'b', label='Validation loss')
plt.title('Training and validation loss')
plt.legend()
 
plt.show()

from keras.preprocessing import image

from keras.models import load_model
import numpy as np
#加载模型
model=load_model('fruit_and_vegetable_100epoch.h5')
#本地图片路径
img_path=("C:/Users/linyicheng/Desktop/base/train/fruit/apple_42.jpg")
img = image.load_img(img_path,target_size=(150,150))
# 将其转换为具有形状的Numpy数组（150、150、3）
img_tensor = image.img_to_array(img)/255
# 将其形态变为(150,150,150,3)的形状
img_tensor = np.expand_dims(img_tensor, axis=0)
#取图片信息
prediction =model.predict(img_tensor)
#输出识别率
print(prediction)
if prediction>0.5:
    print('水果')
else:
    print('蔬菜')
plt.imshow(img)

from PIL import Image
from keras.preprocessing import image
import matplotlib.pyplot as plt
from keras.models import load_model
plt.rcParams['font.family'] = ['SimHei']
figure,ax = plt.subplots(nrows=2, ncols=4, sharex=True, sharey=True,figsize=(24,10))
def convertjpg(jpgfile):
    #将图片缩小到（150，150）的大小
    img=Image.open(jpgfile)
    try:
        new_img=img.resize((150,150),Image.BILINEAR)
        return new_img
    except Exception as e:
        print(e)
snames=[os.path.join(test_vegetable_data,sname)for sname in os.listdir(test_vegetable_data)]
j=0
for i in snames[0:8] :
    j+=1
    img_show=convertjpg(i)
    img_scale = image.img_to_array(img_show)
    img_scale = img_scale.reshape(1,150,150,3) # 将形状转化为(1,150,150,3)
    img_scale = img_scale.astype('float32')/255
    result = model.predict(img_scale)# 预测函数-
    if result>0.5:
        title = '蔬菜'
        label = '正确识别率为：'+str(result)
    else:
        title = '水果'
        label = '正确识别率为：'+str(1-result)
         
    ax = ax.flatten()   # 将子图从多维变成一维
    ax[j-1].imshow(img_show, cmap='Greys', interpolation='nearest')
     # 子图标题
    ax[j-1].set_title(title,fontsize=24)
     # 子图X轴标签
    ax[j-1].set_xlabel(label,fontsize=24)
     # 去掉刻度
    ax[0].set_xticks([])
    ax[0].set_yticks([]) 
plt.show()

figure,ax = plt.subplots(nrows=2, ncols=4, sharex=True, sharey=True,figsize=(24,10))

fnames=[os.path.join(test_fruit_data,fname)for fname in os.listdir(test_fruit_data)]
j=0
for i in fnames[0:8]: 
    j+=1
    img_show=convertjpg(i)
    img_scale = image.img_to_array(img_show)
    img_scale = img_scale.reshape(1,150,150,3) # 将形状转化为(1,150,150,3)
    img_scale = img_scale.astype('float32')/255
    result = model.predict(img_scale)# 预测函数-
    if result>0.5:
        title = '蔬菜'
        label = '正确识别率为：'+ str(result)
    else:
        title = '水果'
        label = '正确识别率为：'+str(1-result)         
    ax = ax.flatten()   # 将子图从多维变成一维
    ax[j-1].imshow(img_show, cmap='Greys', interpolation='nearest')
     # 子图标题
    ax[j-1].set_title(title,fontsize=24)
     # 子图X轴标签
    ax[j-1].set_xlabel(label,fontsize=24)
    # 去掉刻度
    ax[0].set_xticks([])
    ax[0].set_yticks([])

四、总结

此次机器学习过程中参考了minist案例和猫狗识别案例，在数据处理构建数据集目录是出现一些问题，开始思路没有构建好，导致数据处理花费太多时间，以及代码太过冗长，还有就是在导入一些必要的数据库是出现问题如PIL，在网上查明原因说需要更新pillow和PIL的版本，后续应多加练习熟悉所需的各种环境。

这次机器学习中，无论是第一次训练模型还是增强后的模型训练都达到了预期的效果，但是在进行机器学习设计的过程中发现自己对程序设计的思路没有理清楚，导致花费了大量时间改正，以后在进行类似学习前应当将思路理清，再付诸行动。