keras实现图像识别—cat

目的：识别图片中是否有猫
训练集x_train（209,64,64,3）,y_train(209,)
测试集x_test（50,64,64,3），y_test(50,)

设计的神经网络结构图如下：
卷积层——池化层——卷积层——池化层——卷积层——池化层——平铺展开——全连接层——全连接层——全连接层

获取数据

# coding: utf-8
# 2019/7/28 19:23
import h5py
import numpy as np
def get_train():
    f = h5py.File('dataset/train_catvnoncat.h5','r')
    x_train = np.array(f['train_set_x'])#训练集数据  将数据转化为np.array
    y_train = np.array(f['train_set_y'])#训练集标签
    return x_train,y_train
def get_test():
    f = h5py.File('dataset/test_catvnoncat.h5', 'r')
    x_test = np.array(f['test_set_x'])#测试集数据  将数据转化为np.array
    y_test = np.array(f['test_set_y'])#测试集标签
    return x_test,y_test

def preprocess(X):
    #将X标准化，从0-255变成0-1
    X =X / 255
    return X


if __name__ == '__main__':
    x1,y1 = get_train()
    x2,y2 = get_test()
    print(x1.shape,y1.shape)
    print(x2.shape,y2.shape)
    print(np.max(x1))
    print(np.max(y2))

    from matplotlib import  pyplot as plt
    plt.figure()
    for i in range(1,16):
        plt.subplot(3,5,i)
        plt.imshow(x1[i])
        print(y1[i])
    plt.show()

训练测试

# coding: utf-8
# 2019/7/28 19:17
import data
from keras.models import Sequential
from keras.layers import Conv2D,MaxPool2D,Dense,Flatten
if __name__ == '__main__':
    model = Sequential()
    model.add(Conv2D(filters=16,kernel_size=(9,9),strides=1,padding='valid'))
    model.add(MaxPool2D(pool_size=(2,2),strides=2))
    model.add(Conv2D(filters=32,kernel_size=(5,5),strides=1,padding='valid'))
    model.add(MaxPool2D(pool_size=(2,2),strides=2))
    model.add(Conv2D(filters=64,kernel_size=(3,3),strides=1,padding='same'))
    model.add(MaxPool2D(pool_size=(2,2),strides=2))
    model.add(Flatten())
    model.add(Dense(units=256,activation='relu'))
    model.add(Dense(units=128,activation='relu'))
    model.add(Dense(units=1,activation='sigmoid'))

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

    x_train,y_train = data.get_train()
    x_test,y_test = data.get_test()
    x_train = data.preprocess(x_train)
    x_test = data.preprocess(x_test)
    
    model.fit(x_train,y_train,epochs=100)

    result = model.evaluate(x_test,y_test)
    print('测试结果：',result)

结果：可以看到，在训练集上损失值已经极小了，准确率已经达到了1，但是在测试集上准确率只有74%，说明由于训练数据太少，模型存在严重过拟合现象，等我下次再优化一下，添加正则化来解决过拟合问题，比如：L2正则项，Dropout随机失活，Data Augmentation或者Early Stopping等

~~未完待续 o.o

------------------------------------------------------分割线------------------------------------------------------------
第二天，为了解决过拟合问题，我在卷积层和全连接层加入了L2正则项如下：

model = Sequential()
    model.add(Conv2D(filters=16,kernel_size=(9,9),strides=1,padding='valid',kernel_regularizer=regularizers.l2(0.01)))
    model.add(MaxPool2D(pool_size=(2,2),strides=2))
    model.add(Conv2D(filters=32,kernel_size=(5,5),strides=1,padding='valid',kernel_regularizer=regularizers.l2(0.01)))
    model.add(MaxPool2D(pool_size=(2,2),strides=2))
    model.add(Conv2D(filters=64,kernel_size=(3,3),strides=1,padding='same',kernel_regularizer=regularizers.l2(0.01)))
    model.add(MaxPool2D(pool_size=(2,2),strides=2))
    model.add(Flatten())
    model.add(Dense(units=256,activation='relu',kernel_regularizer=regularizers.l2(0.01)))
    model.add(Dense(units=128,activation='relu',kernel_regularizer=regularizers.l2(0.01)))
    model.add(Dense(units=1,activation='sigmoid',kernel_regularizer=regularizers.l2(0.01)))

结果：

相比之前，在测试集上的损失值从1.886下降到了0.890，准确率从0.7399上升到了0.8600，效果确实有所提升。