#使用猫狗数据集进行测试，初始网络为keras VGG16，也可以换为其它网络
# -*- coding: utf-8 -*-
"""
@from:www.linerect.com
@author: duimu
@contact: duimu@qq.com
@Created on: 2019/11/21 9:31
"""
import tensorflow as tf
from keras import applications
from keras.preprocessing.image import ImageDataGenerator
from keras.models import Sequential,Model
from keras.layers import Dropout,Flatten,Dense
from keras import optimizers
import cv2
from PIL import Image
import os
image_width,image_height=256,256
train_data_dir='dogs_cats/train'
validation_data_dir='dogs_cats/validata'
OUT_CATEGORIES=1
nb_train_samples=2000
nb_validation_samples=1000
#loade vgg16
base_model=applications.VGG16(weights="imagenet",include_top=False,input_shape=(image_width,image_height,3))
print("base_model")
base_model.summary()
#freeze the first 15 layers
#base_model.layers[:15].trainable=False
for layer in base_model.layers[:15]:
layer.trainable = False
#define networks
#type 1
"""
top_model=Sequential()
top_model.add(Flatten(input_shape=base_model.output_shape[1:]))
top_model.add(Dense(256,activation='relu'))
top_model.add(Dropout(0.5))
top_model.add(Dense(OUT_CATEGORIES,activation='sigmoid'))
print("top_model")
top_model.summary()
new_model = Model(inputs=base_model.input, outputs=top_model(base_model.output))
"""
#type 2
top_model=base_model.output
top_model=Flatten()(top_model)
top_model=Dense(256,activation='relu')(top_model)
top_model=Dropout(0.5)(top_model)
top_model=Dense(OUT_CATEGORIES,activation='sigmoid')(top_model)
print("new_model")
new_model = Model(inputs=base_model.input, outputs=top_model)
new_model.summary()
#去掉注释，可以进行训练
"""
new_model.compile(optimizer=optimizers.SGD(lr=0.0001, momentum=0.9), loss='binary_crossentropy', metrics=['accuracy'])
#训练数据
batchsize=20
steps=nb_train_samples//batchsize
train_datagen = ImageDataGenerator(1/255.)
train_generator = train_datagen.flow_from_directory(directory=train_data_dir,
target_size=(image_width,image_height),
class_mode='binary',
batch_size=batchsize)
test_datagen = ImageDataGenerator(1/255.)
test_generator = test_datagen.flow_from_directory(directory=train_data_dir,
target_size=(image_width,image_height),
class_mode='binary',
batch_size=batchsize)
new_model.fit_generator(train_generator,
steps_per_epoch=steps,
epochs=5,
validation_data=test_generator,
validation_steps=steps,
verbose=2,
workers=12)
new_model.save_weights("model/迁移学习.dat")
score=new_model.evaluate_generator(test_datagen,nb_validation_samples/batchsize)
s=new_model.predict_generator(test_datagen,nb_validation_samples/batchsize)
"""
#测试训练完成的数据
###-----------------------------------------------------------------------------------------------
#加载训练好的模型
Model.load_weights(new_model,"model/迁移学习.dat")
print("load_model")
new_model.summary()
def file_name(file_dir):
FileList=[]
for root, dirs, files in os.walk(file_dir):
for file in files:
if os.path.splitext(file)[1] == '.jpg':
FileList.append(os.path.join(root, file))
return FileList
#获取文件名称
filenames=file_name('F:/1_Tensorflow/1_Test_Projects/dogs_cats/validata')
pathlen=len('F:/1_Tensorflow/1_Test_Projects/dogs_cats/validata')+1
filenames.sort(key = lambda x: int(x[pathlen:-4]))
for file in filenames :
#BGR
image = cv2.imread(os.path.join(file))
size = (image_width, image_height)
image = cv2.resize(image, size)
#转化RGB
img2 = image[:, :, :: -1]
input = tf.reshape(img2, shape=[-1, image_width, image_height, 3])
result = new_model.predict(input, steps=1)
print(result,file)
