1 前言

Hi，大家好，这里是丹成学长，今天做一个 基于深度学习的水果识别demo

这是一个较为新颖的竞赛课题方向，学长非常推荐！

🧿 更多资料, 项目分享：

https://gitee.com/dancheng-senior/postgraduate

2 开发简介

深度学习作为机器学习领域内新兴并且蓬勃发展的一门学科， 它不仅改变着传统的机器学习方法， 也影响着我们对人类感知的理解，
已经在图像识别和语音识别等领域取得广泛的应用。 因此， 本文在深入研究深度学习理论的基础上， 将深度学习应用到水果图像识别中，
以此来提高了水果图像的识别性能。

3 识别原理

3.1 传统图像识别原理

传统的水果图像识别系统的一般过程如下图所示，主要工作集中在图像预处理和特征提取阶段。

在大多数的识别任务中， 实验所用图像往往是在严格限定的环境中采集的， 消除了外界环境对图像的影响。 但是实际环境中图像易受到光照变化、 水果反光、
遮挡等因素的影响， 这在不同程度上影响着水果图像的识别准确率。

在传统的水果图像识别系统中， 通常是对水果的纹理、 颜色、 形状等特征进行提取和识别。

3.2 深度学习水果识别

CNN 是一种专门为识别二维特征而设计的多层神经网络， 它的结构如下图所示，这种结构对平移、 缩放、 旋转等变形具有高度的不变性。

学长本次采用的 CNN 架构如图：

4 数据集

• 数据库分为训练集(train)和测试集(test)两部分

• 训练集包含四类apple,orange,banana,mixed(多种水果混合)四类237张图片；测试集包含每类图片各两张。图片集如下图所示。

• 图片类别可由图片名称中提取。

训练集图片预览

测试集预览

数据集目录结构

5 部分关键代码

5.1 处理训练集的数据结构

import os
import pandas as pd    

train_dir = './Training/'
test_dir = './Test/'
fruits = []
fruits_image = []

for i in os.listdir(train_dir):
    for image_filename in os.listdir(train_dir + i):
        fruits.append(i) # name of the fruit
        fruits_image.append(i + '/' + image_filename)
train_fruits = pd.DataFrame(fruits, columns=["Fruits"])
train_fruits["Fruits Image"] = fruits_image

print(train_fruits)

5.2 模型网络结构

    
​    import matplotlib.pyplot as plt
​    import seaborn as sns
​    from keras.preprocessing.image import ImageDataGenerator, img_to_array, load_img
​    from glob import glob
​    from keras.models import Sequential
​    from keras.layers import Conv2D, MaxPooling2D, Activation, Dropout, Flatten, Dense
​    img = load_img(train_dir + "Cantaloupe 1/r_234_100.jpg")
​    plt.imshow(img)
​    plt.axis("off")
​    plt.show()
​    

    array_image = img_to_array(img)
    
    # shape (100,100)
    print("Image Shape --> ", array_image.shape)
    
    # 131个类目
    fruitCountUnique = glob(train_dir + '/*' )
    numberOfClass = len(fruitCountUnique)
    print("How many different fruits are there --> ",numberOfClass)
    
    # 构建模型
    model = Sequential()
    model.add(Conv2D(32,(3,3),input_shape = array_image.shape))
    model.add(Activation("relu"))
    model.add(MaxPooling2D())
    model.add(Conv2D(32,(3,3)))
    model.add(Activation("relu"))
    model.add(MaxPooling2D())
    model.add(Conv2D(64,(3,3)))
    model.add(Activation("relu"))
    model.add(MaxPooling2D())
    model.add(Flatten())
    model.add(Dense(1024))
    model.add(Activation("relu"))
    model.add(Dropout(0.5))
    
    # 区分131类
    model.add(Dense(numberOfClass)) # output
    model.add(Activation("softmax"))
    model.compile(loss = "categorical_crossentropy",
    
                  optimizer = "rmsprop",
    
                  metrics = ["accuracy"])
    
    print("Target Size --> ", array_image.shape[:2])


## 

5.3 训练模型

    
​    train_datagen = ImageDataGenerator(rescale= 1./255,
​                                       shear_range = 0.3,
​                                       horizontal_flip=True,
​                                       zoom_range = 0.3)
​    

    test_datagen = ImageDataGenerator(rescale= 1./255)
    epochs = 100
    batch_size = 32
    train_generator = train_datagen.flow_from_directory(
                    train_dir,
                    target_size= array_image.shape[:2],
                    batch_size = batch_size,
                    color_mode= "rgb",
                    class_mode= "categorical")
    
    test_generator = test_datagen.flow_from_directory(
                    test_dir,
                    target_size= array_image.shape[:2],
                    batch_size = batch_size,
                    color_mode= "rgb",
                    class_mode= "categorical")
    
    for data_batch, labels_batch in train_generator:
        print("data_batch shape --> ",data_batch.shape)
        print("labels_batch shape --> ",labels_batch.shape)
        break
    
    hist = model.fit_generator(
            generator = train_generator,
            steps_per_epoch = 1600 // batch_size,
            epochs=epochs,
            validation_data = test_generator,
            validation_steps = 800 // batch_size)
    
    #保存模型 model_fruits.h5
    model.save('model_fruits.h5')

顺便输出训练曲线

​

    #展示损失模型结果
​    plt.figure()
​    plt.plot(hist.history["loss"],label = "Train Loss", color = "black")
​    plt.plot(hist.history["val_loss"],label = "Validation Loss", color = "darkred", linestyle="dashed",markeredgecolor = "purple", markeredgewidth = 2)
​    plt.title("Model Loss", color = "darkred", size = 13)
​    plt.legend()
​    plt.show()
​    

    #展示精确模型结果
    plt.figure()
    plt.plot(hist.history["accuracy"],label = "Train Accuracy", color = "black")
    plt.plot(hist.history["val_accuracy"],label = "Validation Accuracy", color = "darkred", linestyle="dashed",markeredgecolor = "purple", markeredgewidth = 2)
    plt.title("Model Accuracy", color = "darkred", size = 13)
    plt.legend()
    plt.show()


![在这里插入图片描述](https://img-blog.csdnimg.cn/686ace7db27c4145837ec2e09e8ad917.png?x-oss-process=image/watermark,type_ZHJvaWRzYW5zZmFsbGJhY2s,shadow_50,text_Q1NETiBARGFuQ2hlbmctc3R1ZGlv,size_17,color_FFFFFF,t_70,g_se,x_16)

6 识别效果

from tensorflow.keras.models import load_model
import os
import pandas as pd
from keras.preprocessing.image import ImageDataGenerator,img_to_array, load_img
import cv2,matplotlib.pyplot as plt,numpy as np
from keras.preprocessing import image

train_datagen = ImageDataGenerator(rescale= 1./255,
                                    shear_range = 0.3,
                                    horizontal_flip=True,
                                    zoom_range = 0.3)

model = load_model('model_fruits.h5')
batch_size = 32
img = load_img("./Test/Apricot/3_100.jpg",target_size=(100,100))
plt.imshow(img)
plt.show()

array_image = img_to_array(img)
array_image = array_image * 1./255
x = np.expand_dims(array_image, axis=0)
images = np.vstack([x])
classes = model.predict_classes(images, batch_size=10)
print(classes)
train_dir = './Training/'

train_generator = train_datagen.flow_from_directory(
        train_dir,
        target_size= array_image.shape[:2],
        batch_size = batch_size,
        color_mode= "rgb",
        class_mode= "categorical”)
print(train_generator.class_indices)

​

​

    fig = plt.figure(figsize=(16, 16))
    axes = []
    files = []
    predictions = []
    true_labels = []
    rows = 5
    cols = 2
# 随机选择几个图片
def getRandomImage(path, img_width, img_height):
    """function loads a random image from a random folder in our test path"""
    folders = list(filter(lambda x: os.path.isdir(os.path.join(path, x)), os.listdir(path)))
    random_directory = np.random.randint(0, len(folders))
    path_class = folders[random_directory]
    file_path = os.path.join(path, path_class)
    file_names = [f for f in os.listdir(file_path) if os.path.isfile(os.path.join(file_path, f))]
    random_file_index = np.random.randint(0, len(file_names))
    image_name = file_names[random_file_index]
    final_path = os.path.join(file_path, image_name)
    return image.load_img(final_path, target_size = (img_width, img_height)), final_path, path_class

def draw_test(name, pred, im, true_label):
    BLACK = [0, 0, 0]
    expanded_image = cv2.copyMakeBorder(im, 160, 0, 0, 300, cv2.BORDER_CONSTANT, value=BLACK)
    cv2.putText(expanded_image, "predicted: " + pred, (20, 60), cv2.FONT_HERSHEY_SIMPLEX,
        0.85, (255, 0, 0), 2)
    cv2.putText(expanded_image, "true: " + true_label, (20, 120), cv2.FONT_HERSHEY_SIMPLEX,
        0.85, (0, 255, 0), 2)
    return expanded_image
IMG_ROWS, IMG_COLS = 100, 100

# predicting images
for i in range(0, 10):
    path = "./Test"
    img, final_path, true_label = getRandomImage(path, IMG_ROWS, IMG_COLS)
    files.append(final_path)
    true_labels.append(true_label)
    x = image.img_to_array(img)
    x = x * 1./255
    x = np.expand_dims(x, axis=0)
    images = np.vstack([x])
    classes = model.predict_classes(images, batch_size=10)
    predictions.append(classes)

class_labels = train_generator.class_indices
class_labels = {v: k for k, v in class_labels.items()}
class_list = list(class_labels.values())

for i in range(0, len(files)):
    image = cv2.imread(files[i])
    image = draw_test("Prediction", class_labels[predictions[i][0]], image, true_labels[i])
    axes.append(fig.add_subplot(rows, cols, i+1))
    plt.imshow(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
    plt.grid(False)
    plt.axis('off')
plt.show()

7 最后

🧿 更多资料, 项目分享：

https://gitee.com/dancheng-senior/postgraduate