深度学习实战（1）：花的分类任务

本站内容均来自兴趣收集,如不慎侵害的您的相关权益,请留言告知,我们将尽快删除.谢谢.

写在前面：

实验目的：通过建立Alexnet神经网络建立模型并根据训练数据来训练模型 以达到可以将一张花的类别进行分类

Python版本：Python3

IDE：VSCode

系统：MacOS

数据集以及代码的资源放在文章末尾了 有需要请自取～

目录

训练模型代码 （附有注释）

训练结果 Accuracy展示

前言

本文仅作为学习训练 不涉及任何商业用途 如有错误或不足之处还请指出

数据集

数据集一共有五种花的类别 但本次实验模型仅用了rose和sunflower两种类别进行分类测试

五种花的类别：

Rose：

Sunflower：

训练模型代码 （附有注释）

import os , glob
from sklearn.model_selection import train_test_split
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras import layers
import matplotlib.pyplot as plt
# 变量
resize = 224 # 图片尺寸参数
epochs = 8 # 迭代次数
batch_size = 5 # 每次训练多少张
#——————————————————————————————————————————————————————————————————————————————————
# 训练集路径
train_data_path = '/Users/liqun/Desktop/KS/MyPython/DataSet/flowers/Training'
# 玫瑰花文件夹路径
rose_path = os.path.join(train_data_path,'rose')
# 太阳花文件夹路径
sunflower_path = os.path.join(train_data_path,'sunflower')
# 将文件夹内的图片读取出来
fpath_rose = [os.path.abspath(fp) for fp in glob.glob(os.path.join(rose_path,'*.jpg'))]
fpath_sunflower = [os.path.abspath(fp) for fp in glob.glob(os.path.join(sunflower_path,'*.jpg'))]
#文件数量
num_rose = len(fpath_rose)
num_sunflower = len(fpath_sunflower)
# 设置标签
label_rose = [0] * num_rose
label_sunflower = [1] * num_sunflower
# 展示
print('rose:   ', num_rose)
print('sunflower: ', num_sunflower)
# 划分为多少验证集
RATIO_TEST = 0.1
num_rose_test = int(num_rose * RATIO_TEST)
num_sunflower_test = int(num_sunflower * RATIO_TEST)
# train
fpath_train = fpath_rose[num_rose_test:] + fpath_sunflower[num_sunflower_test:]
label_train = label_rose[num_rose_test:] + label_sunflower[num_sunflower_test:]
# validation
fpath_vali = fpath_rose[:num_rose_test] + fpath_sunflower[:num_sunflower_test]
label_vali = label_rose[:num_rose_test] + label_sunflower[:num_sunflower_test]
num_train = len(fpath_train)
num_vali = len(fpath_vali)
# 展示
print('num_train:   ', num_train)
print('num_label: ', num_vali)
# 预处理函数
def preproc(fpath, label):
    image_byte = tf.io.read_file(fpath) # 读取文件
    image = tf.io.decode_image(image_byte) # 检测图像是否为BMP,GIF,JPEG或PNG,并执行相应的操作将输入字节string转换为类型uint8的Tensor
    image_resize = tf.image.resize_with_pad(image, 224, 224) #缩放到224*224
    image_norm = tf.cast(image_resize, tf.float32) / 255. #归一化
    
    label_onehot = tf.one_hot(label, 2)
    
    return image_norm, label_onehot
dataset_train = tf.data.Dataset.from_tensor_slices((fpath_train, label_train)) #将数据进行预处理
dataset_train = dataset_train.shuffle(num_train).repeat()  #打乱顺序
dataset_train = dataset_train.map(preproc, num_parallel_calls=tf.data.experimental.AUTOTUNE)
dataset_train = dataset_train.batch(batch_size).prefetch(tf.data.experimental.AUTOTUNE) #一批次处理多少份
dataset_vali = tf.data.Dataset.from_tensor_slices((fpath_vali, label_vali))
dataset_vali = dataset_vali.shuffle(num_vali).repeat()
dataset_vali = dataset_vali.map(preproc, num_parallel_calls=tf.data.experimental.AUTOTUNE)
dataset_vali = dataset_vali.batch(batch_size).prefetch(tf.data.experimental.AUTOTUNE)
#——————————————————————————————————————————————————————————————————————————————————
# 建立模型 卷积神经网络
model = tf.keras.Sequential(name='Alexnet')
# 第一层
model.add(layers.Conv2D(filters=96, kernel_size=(11,11),
                    strides=(4,4), padding='valid',
                    input_shape=(resize,resize,3),
                    activation='relu'))
model.add(layers.BatchNormalization())
# 第一层池化层：最大池化层
model.add(layers.MaxPooling2D(pool_size=(3,3), 
                    strides=(2,2), 
                    padding='valid'))
#第二层
model.add(layers.Conv2D(filters=256, kernel_size=(5,5), 
                    strides=(1,1), padding='same', 
                    activation='relu'))
model.add(layers.BatchNormalization())
#第二层池化层
model.add(layers.MaxPooling2D(pool_size=(3,3), 
                    strides=(2,2), 
                    padding='valid'))
#第三层
model.add(layers.Conv2D(filters=384, kernel_size=(3,3), 
                    strides=(1,1), padding='same', 
                    activation='relu'))
#第四层
model.add(layers.Conv2D(filters=384, kernel_size=(3,3), 
                    strides=(1,1), padding='same', 
                    activation='relu'))
#第五层
model.add(layers.Conv2D(filters=256, kernel_size=(3,3), 
                    strides=(1,1), padding='same', 
                    activation='relu'))
#池化层
model.add(layers.MaxPooling2D(pool_size=(3,3), 
                    strides=(2,2), padding='valid'))
#第6，7，8层
model.add(layers.Flatten())
model.add(layers.Dense(4096, activation='relu'))
model.add(layers.Dropout(0.5))
model.add(layers.Dense(4096, activation='relu'))
model.add(layers.Dropout(0.5))
model.add(layers.Dense(1000, activation='relu'))
model.add(layers.Dropout(0.5))
# Output Layer
model.add(layers.Dense(2, activation='softmax'))
# Training 优化器 随机梯度下降算法
model.compile(loss='categorical_crossentropy',
                optimizer='sgd',           #梯度下降法
                metrics=['accuracy'])
history = model.fit(dataset_train,
            steps_per_epoch = num_train//batch_size,
            epochs = epochs,         #迭代次数
            validation_data = dataset_vali,
            validation_steps = num_vali//batch_size,
            verbose = 1)
# 评分标准
scores_train = model.evaluate(dataset_train, steps=num_train//batch_size, verbose=1)
print(scores_train)
scores_vali = model.evaluate(dataset_vali, steps=num_vali//batch_size, verbose=1)
print(scores_vali)
#保存模型
model.save('/Users/liqun/Desktop/KS/MyPython/project/flowerModel.h5')
'''
history对象的history内容（history.history）是字典类型，
键的内容受metrics的设置影响，值的长度与epochs值一致。
'''
history_dict = history.history
train_loss = history_dict['loss']
train_accuracy = history_dict['accuracy']
val_loss = history_dict['val_loss']
val_accuracy = history_dict['val_accuracy']
# Draw loss
plt.figure()
plt.plot(range(epochs), train_loss, label='train_loss')
plt.plot(range(epochs), val_loss, label='val_loss')
plt.legend()
plt.xlabel('epochs')
plt.ylabel('loss')
# Draw accuracy
plt.figure()
plt.plot(range(epochs), train_accuracy, label='train_accuracy')
plt.plot(range(epochs), val_accuracy, label='val_accuracy')
plt.legend()
plt.xlabel('epochs')
plt.ylabel('accuracy')
# Display
plt.show()
print('Train has finished')

训练集数据量展示

训练迭代过程展示

训练结果 Accuracy展示

训练结果 Loss展示

测试集

预测结果代码

import cv2
from tensorflow.keras.models import load_model
resize = 224
label = ('rose', 'sunflower')
image = cv2.resize(cv2.imread('/Users/liqun/Desktop/KS/MyPython/DataSet/flowers/Training/sunflower/23286304156_3635f7de05.jpg'),(resize,resize))
image = image.astype("float") / 255.0 # 归一化
image = image.reshape((1, image.shape[0], image.shape[1], image.shape[2]))
# 加载模型
model = load_model('/Users/liqun/Desktop/KS/MyPython/project/flowerModel.h5')
predict = model.predict(image)
i = predict.argmax(axis=1)[0]
# 展示结果
print('——————————————————————')
print('Predict result')
print(label[i],':',max(predict[0])*100,'%')

预测结果展示

结语

模型到这里就训练并检测完毕了 如有需要的小伙伴可以下载下方的数据集测试集及源代码

链接: https://pan.baidu.com/s/1OJfwcF1PvX9qkZwT7MXd_Q?pwd=i0bt 提取码: i0bt