目的
突发奇想想会认为下面这张图片究竟是玛丽莲梦露还是爱因斯坦，主要目的顺便实践练习《Python深度学习》书中的例子，只采用了很小批量的数据，也没有深究如何提高正确率，解决过拟合的问题。详细可以参见《python深度学习》第五章前两节。
数据准备
从百度图片中找到了风格各异的爱因斯坦的图片，直接采用下载整个网页的方式获取图片。选的量不多，100张作为训练，25张用于验证。本来是留有测试的数据，不小心删掉了就跳过在新数据上测试的步骤。（数据量太小也是一个严重的问题）
手动删掉一些不合适的图片，分别放到train和validation文件夹下的E，M两个文件中。
构建网络
建立序列模型，采用这个网络是因为之前在一个SAR图像的识别中表现优异，预测准确率达到96%以上（尽管并不能说明它在区分爱因斯坦和玛丽莲梦露也能表现得很好）
import osfrom keras import layersfrom keras import modelsmodel = models.Sequential()model.add(layers.Conv2D(64, (3, 3), activation='relu',input_shape=(88, 88,3)))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.Dropout(0.5))model.add(layers.Dense(512, activation='relu'))model.add(layers.Dense(1, activation='sigmoid'))model.summary()_________________________________________________________________Layer (type) Output Shape Param # =================================================================conv2d_1 (Conv2D) (None, 86, 86, 64) 1792_________________________________________________________________max_pooling2d_1 (MaxPooling2 (None, 43, 43, 64) 0 _________________________________________________________________conv2d_2 (Conv2D) (None, 41, 41, 64) 36928_________________________________________________________________max_pooling2d_2 (MaxPooling2 (None, 20, 20, 64) 0 _________________________________________________________________conv2d_3 (Conv2D) (None, 18, 18, 128) 73856_________________________________________________________________max_pooling2d_3 (MaxPooling2 (None, 9, 9, 128) 0 _________________________________________________________________conv2d_4 (Conv2D) (None, 7, 7, 128) 147584 _________________________________________________________________max_pooling2d_4 (MaxPooling2 (None, 3, 3, 128) 0 _________________________________________________________________flatten_1 (Flatten)(None, 1152) 0 _________________________________________________________________dropout_1 (Dropout)(None, 1152) 0 _________________________________________________________________dense_1 (Dense) (None, 512)590336 _________________________________________________________________dropout_2 (Dropout)(None, 512)0 _________________________________________________________________dense_2 (Dense) (None, 512)262656 _________________________________________________________________dense_3 (Dense) (None, 1) 513 =================================================================Total params: 1,113,665Trainable params: 1,113,665Non-trainable params: 0
读入图片并训练
base_dir = r'dir\Einstein'train_dir = os.path.join(base_dir, 'train')validation_dir = os.path.join(base_dir, 'v')from keras import pile(loss='binary_crossentropy',optimizer=optimizers.RMSprop(lr=1e-4),metrics=['acc'])from keras.preprocessing.image import ImageDataGeneratortrain_datagen = ImageDataGenerator(rescale=1. / 255)test_datagen = ImageDataGenerator(rescale=1. / 255)train_generator = train_datagen.flow_from_directory(train_dir,target_size=(88,88),batch_size=20,class_mode='binary')validation_generator = test_datagen.flow_from_directory(validation_dir,target_size=(88, 88),batch_size=20,class_mode='binary')history = model.fit_generator(train_generator,steps_per_epoch=128,epochs=20,validation_data=validation_generator,validation_steps=50)#保存模型model.save('EM.h5')
训练结果
绘制性能曲线
import matplotlib.pyplot as pltacc = 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()
l最终结果：
oss: 0.0050 - acc: 0.9996 -
val_loss: 2.4635 - val_acc: 0.6614
结果显然过拟合了，预测的正确率只有66%。但不妨碍用于用的预测。
预测
from keras.preprocessing import imageimport matplotlib.image as mpimgfrom keras import modelsimport numpy as npimg = image.load_img(r'dir\Einstein\EM.jpg',target_size=(88,88,3))img = np.array(img)img = img/255model = models.load_model(r'dir\Einstein\EM.h5')img = img.reshape(1,88,88,3)pre = model.predict(img)print('预测结果：'，pre)预测结果： [[0.00376787]]
Keras添加的标签是E（爱因斯坦）文件夹中的为0，M（玛丽莲梦露）为1。通过网络最后的sigmoid单元，输出值为0.00376787，这个神经网络十分倾向于认为这张图片是爱因斯坦。
尝试了很多种不同的结构（数据量小训练也很快），验证集的正确率一直在70%左右，仅有一次认为该图片是玛丽莲梦露，其余结果都认为这张图片是爱因斯坦。
结论
在搭建的这样的简单的网络下，更倾向于认为这种图片里的人是爱因斯坦。
不足之处
样本太少过拟合，验证集的识别正确率不高
参考资料
《python深度学习》
