def main():
fnet = SiameseFaceNet()
model_dir_path = './models'
image_dir_path = "./data/images"
fnet.load_model(model_dir_path)
database = dict()
database["danielle"] = [fnet.img_to_encoding(image_dir_path + "/danielle.png")]
database["younes"] = [fnet.img_to_encoding(image_dir_path + "/younes.jpg")]
database["tian"] = [fnet.img_to_encoding(image_dir_path + "/tian.jpg")]
database["andrew"] = [fnet.img_to_encoding(image_dir_path + "/andrew.jpg")]
database["kian"] = [fnet.img_to_encoding(image_dir_path + "/kian.jpg")]
database["dan"] = [fnet.img_to_encoding(image_dir_path + "/dan.jpg")]
database["sebastiano"] = [fnet.img_to_encoding(image_dir_path + "/sebastiano.jpg")]
database["bertrand"] = [fnet.img_to_encoding(image_dir_path + "/bertrand.jpg")]
database["kevin"] = [fnet.img_to_encoding(image_dir_path + "/kevin.jpg")]
database["felix"] = [fnet.img_to_encoding(image_dir_path + "/felix.jpg")]
database["benoit"] = [fnet.img_to_encoding(image_dir_path + "/benoit.jpg")]
database["arnaud"] = [fnet.img_to_encoding(image_dir_path + "/arnaud.jpg")]
database["deepak"] = [fnet.img_to_encoding(image_dir_path + "/deepak.jpg")]
database["momo"] = [fnet.img_to_encoding(image_dir_path + "/momo.jpg")]
fnet.verify(image_dir_path + "/camera_0.jpg", "younes", database)
fnet.verify(image_dir_path + "/camera_2.jpg", "kian", database)
# fnet.verify(image_dir_path + "/test_deepak_1.jpg", "deepak", database)
# fnet.verify(image_dir_path + "/test_momo_1.jpg", "momo", database)
# fnet.verify(image_dir_path + "/test_momo.jpg", "deepak", database)
fnet.who_is_it(image_dir_path + "/camera_0.jpg", database)
fnet.who_is_it(image_dir_path + "/younes.jpg", database)
def main():
fnet = SiameseFaceNet()
fnet.vgg16_include_top = True
model_dir_path = './models_1'
image_dir_path = "./data/images"
database = dict()
database["aipengfei"] = [fnet.img_to_encoding(image_dir_path + "/aipengfei.png")]
database["anyaru"] = [fnet.img_to_encoding(image_dir_path + "/anyaru.png")]
database["baozhiqian"] = [fnet.img_to_encoding(image_dir_path + "/baozhiqian.png")]
# database["danielle"] = [fnet.img_to_encoding(image_dir_path + "/danielle.png")]
# database["younes"] = [fnet.img_to_encoding(image_dir_path + "/younes.jpg")]
# database["tian"] = [fnet.img_to_encoding(image_dir_path + "/tian.jpg")]
# database["andrew"] = [fnet.img_to_encoding(image_dir_path + "/andrew.jpg")]
# database["kian"] = [fnet.img_to_encoding(image_dir_path + "/kian.jpg")]
# database["dan"] = [fnet.img_to_encoding(image_dir_path + "/dan.jpg")]
# database["sebastiano"] = [fnet.img_to_encoding(image_dir_path + "/sebastiano.jpg")]
# database["bertrand"] = [fnet.img_to_encoding(image_dir_path + "/bertrand.jpg")]
# database["kevin"] = [fnet.img_to_encoding(image_dir_path + "/kevin.jpg")]
# database["felix"] = [fnet.img_to_encoding(image_dir_path + "/felix.jpg")]
# database["benoit"] = [fnet.img_to_encoding(image_dir_path + "/benoit.jpg")]
# database["arnaud"] = [fnet.img_to_encoding(image_dir_path + "/arnaud.jpg")]
fnet.fit(database=database, model_dir_path=model_dir_path)
def predict(self, database):
fnet = SiameseFaceNet()
model_dir_path = './models_1'
image_dir_path = "./data/images"
fnet.load_model(model_dir_path)
print("-------------------------")
# fnet.verify(image_dir_path + "/001.png", "aipengfei", database)
# fnet.verify(image_dir_path + "/002.png", "aipengfei", database)
# fnet.who_is_it(image_dir_path + "/001.png", database)
#fnet.who_is_it(image_dir_path + "/002.png", database)
fnet.who_is_it("data/5test.jpg", database)
def main():
fnet = SiameseFaceNet()
model_dir_path = './model'
image_dir_path = "./data/test/align"
fnet.load_model(model_dir_path)
database1 = dict() #dictionary
#fnet.who_is_it(image_dir_path + "/younes.jpg", database)
database1["chenaifang"] = [
fnet.img_to_encoding(image_dir_path + "/bailu1.jpg")
]
fnet.verify(image_dir_path + "/wangtiange1.jpg", "chenaifang", database1)
def predict(self):
fnet = SiameseFaceNet()
model_dir_path = './models_1'
image_dir_path = "./data/images"
fnet.load_model(model_dir_path)
database = dict()
database["aipengfei"] = [fnet.img_to_encoding(image_dir_path + "/aipengfei.png")]
database["anyaru"] = [fnet.img_to_encoding(image_dir_path + "/anyaru.png")]
database["baozhiqian"] = [fnet.img_to_encoding(image_dir_path + "/baozhiqian.png")]
print("-------------------------")
fnet.verify(image_dir_path + "/001.png", "aipengfei", database)
fnet.verify(image_dir_path + "/002.png", "aipengfei", database)
fnet.who_is_it(image_dir_path + "/001.png", database)
fnet.who_is_it(image_dir_path + "/002.png", database)
fnet.who_is_it(image_dir_path + "/003.png", database)
def fileName(self):
fnet = SiameseFaceNet()
model_dir_path = './models_1'
image_dir_path = "./data/images"
fnet.load_model(model_dir_path)
database = dict()
rootdir = 'data/stu'
list = os.listdir(rootdir) # 列出文件夹下所有的目录与文件
for i in range(0, len(list)):
path = rootdir + "/" + list[i]
listname = os.listdir(path)
for i in range(0, len(listname)):
print(path + '/' + listname[i])
database[listname[i].split('.')[0]] = [
fnet.img_to_encoding(path + '/' + listname[i])
]
#print (path+'/'+listname[i])
return database
def fileName(self):
fnet = SiameseFaceNet()
model_dir_path = './models_2'
fnet.load_model(model_dir_path)
database = dict()
rootdir = 'data/stu100'
list = os.listdir(rootdir) # 列出文件夹下所有的目录与文件
for index in range(0, len(list)):
path = rootdir+"/"+list[index]
listname = os.listdir(path)
for i in range(0, len(listname)):
print(path+'/'+listname[i])
database[list[index]] = [fnet.img_to_encoding(path+'/'+listname[i])]
#print (path+'/'+listname[i])
output = open('myfile.pkl', 'wb')
pickle.dump(database, output)
output.close()
return database
def predict(self,database):
fnet = SiameseFaceNet()
model_dir_path = './models_2'
fnet.load_model(model_dir_path)
print("-------------------------")
# fnet.verify(image_dir_path + "/001.png", "aipengfei", database)
# fnet.verify(image_dir_path + "/002.png", "aipengfei", database)
# fnet.who_is_it(image_dir_path + "/001.png", database)
#fnet.who_is_it(image_dir_path + "/002.png", database)
fnet.who_is_it("data/test/chencongcong.png", database)
fnet.who_is_it("data/test/chendandan.png", database)
# fnet.who_is_it("data/test/chendaocheng.png", database)
# fnet.who_is_it("data/test/chenfuyuan.png", database)
# fnet.who_is_it("data/test/chenge.png", database)
# fnet.who_is_it("data/test/chenguang.png", database)
# fnet.who_is_it("data/test/chenguangwei.png", database)
# fnet.who_is_it("data/test/chenguoyan.png", database)
# fnet.who_is_it("data/test/chenhaiyan.png", database)
# fnet.who_is_it("data/test/chenhao.png", database)
def main():
fnet = SiameseFaceNet()
'''
因为VGGNET最后有三个全连接层, 因此,这个选项表示是否需要最上面的三个全连接层.
一般网络最后都会有全连接层, 最后一个全连接层更是设定了分类的个数, loss的计算方法,
并架设了一个概率转换函数(soft max). 其实soft max的作用就是将输出转换为各类别的概率,并计算loss.
可以这么说, 最上面三层是用来进行分类的, 其余层是用来进行特征提取的.
因此如果include_top=False,也就表示这个网络只能进行特征提取. 不能在进行新的训练或者在已有权重上fine-tune.
'''
#fnet.vgg16_include_top = True #包含最上层的全连接层
#fnet.threshold=0.7 #阈值，在验证的时候作为是否是同一个人的标准
model_dir_path = './model'
#image_dir_path = "./data/dlib-align-images"
#在测试
image_dir_path = "./data/dlib-align-images"
database = dict()
#单人单张照片时
'''
for filename in os.listdir(r"./data/dlib-align-images"):
list=filename.split(".")
#print(list[0])
database[list[0]]=[fnet.img_to_encoding(image_dir_path+"/"+filename)]
'''
'''#打算将模型整合起来
for filename in os.listdir(r"./data/dlib-align-images-test"):
pic_list = []#key是人名，value是两张图片的名字
for picname in os.listdir(r"./data/dlib-align-images-test/" + filename):
pic_list.append(fnet.img_to_encoding(image_dir_path + "/" + filename + "/" + picname))
#pic_list.append(image_dir_path + "/" + filename + "/" + picname)
database_name[filename] = pic_list
fnet.fit(database1=database_name, model_dir_path=model_dir_path)
'''
#'''
#i=0
#j=0
for filename in os.listdir(r"./data/dlib-align-images"):
pic_list = []
for picname in os.listdir(r"./data/dlib-align-images/" + filename):
pic_list.append(
fnet.img_to_encoding(image_dir_path + "/" + filename + "/" +
picname))
database[filename] = pic_list
np.save('picture_encoding.npy', database)
#'''
read_dictionary = np.load("picture_encoding.npy").item()
#database=read_dictionary
#fnet.fit(database=database, model_dir_path=model_dir_path)
#i=i+1
#if i is 17:
# j = j + 1
# np.save('picture_encoding_qqq.npy'+str(j), database)
# i = 0
# database=dict()
#将字典保存到文件中
#'''
#'''
#读取文件中的内容，将内容赋给字典
#read_dictionary1 = np.load("picture_encoding1.npy").item()
#read_dictionary2 = np.load("picture_encoding2.npy").item()
#database=read_dictionary1.copy()
#database.update(read_dictionary2
#database=dict(read_dictionary1.items()+read_dictionary2.items())
#print(database)
#read_dictionary=database
#random.shuffle(read_dictionary)
database = read_dictionary
#database是人名；
fnet.fit(database=database, model_dir_path=model_dir_path)
def createAug_1():
rootdir = 'data/stu01'
list = os.listdir(rootdir) # 列出文件夹下所有的目录与文件
count = 0
for index in range(0, len(list)):
#print (count)
count = count + 1
path = rootdir + "/" + list[index]
listname = os.listdir(path)
if len(listname) == 0:
print(list[index])
#os.makedirs( 'data/stu01/'+list[index])
for i in range(0, len(listname)):
#print(path + '/' + listname[i])
imgpath = path + '/' + listname[0]
img = cv2.imread(imgpath) # 这是一个PIL图像
#img =cv2.cvtColor( np.array(img), cv2.COLOR_BGR2RGB)
#cv2.imwrite('data/stu01/' +list[index]+'/'+ listname[0], img)
break
if __name__ == '__main__':
#createAug_1()
fnet = SiameseFaceNet()
fnet.vgg16_include_top = True
saveTrainPkl(fnet)
main(fnet)
createAug()