gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) #得到灰度图像
faces = face_cascade.detectMultiScale(gray) #使用opencv自带库寻找照片中是否有人脸
if (len(faces) > 0): #返回值大于0 存在人脸
face_re = True
else:
face_re = False
time.sleep(2)
print face_re
if (face_re): #如何人脸存在
res = api.detect(api_key=API_KEY,
api_secret=API_SECRET,
image_file=File(img_re))
print_result(printFuctionTitle("人脸检测"), res)
# 人脸比对：https://console.faceplusplus.com.cn/documents/4887586
compare_res = api.compare(api_key=API_KEY,
api_secret=API_SECRET,
image_file1=File(img2_re),
image_file2=File(img_re))
#compare_res = api.compare(image_file1=File(face_search_img), image_file2=File(face_search_img))
print_result(printFuctionTitle("compare"), compare_res)
print compare_res.confidence #打印俩张照片的置信度 越高则越相似
# 人脸搜索：https://console.faceplusplus.com.cn/documents/4888381
# 人脸搜索步骤
# 1,创建faceSet:用于存储人脸信息(face_token)
# 2,向faceSet中添加人脸信息(face_token)
# 3，开始搜索
# 删除无用的人脸库，这里删除了，如果在项目中请注意是否要删除
# api.faceset.delete(outer_id='faceplusplus', check_empty=0)
# # 1.创建一个faceSet
# ret = api.faceset.create(outer_id='faceplusplus')
#
class Analyze(object):
def __init__(self,
path_to_analyze_config,
output_dir_name="analyzed_result",
wait_time=None):
self.path_to_analyze_config = path_to_analyze_config
with open(self.path_to_analyze_config, "r") as f:
self.config = yaml.load(f, Loader=yaml.SafeLoader)
self.output_dir_name = output_dir_name
self.wait_time = wait_time #自然数で指定する必要がある．
if self.config["output_json_name"] is None:
self.output_json_name = "output"
else:
self.output_json_name = self.config["output_json_name"]
self.api = API(API_KEY=self.config["API_KEY"],
API_SECRET=self.config["API_SECRET"])
del self.config["API_KEY"], self.config["API_SECRET"]
def get_dataset(self,
serch_config_name="config_camera.yaml",
path_to_image=None):
if path_to_image is None:
path_to_image = self.config["target_dir_name"]
if serch_config_name in os.listdir(path_to_image):
self.abs_path = os.path.abspath(path_to_image)
path_to_camera_config = os.path.join(
*[self.abs_path, serch_config_name])
with open(path_to_camera_config, "r") as f:
self.config_camera = yaml.load(f, Loader=yaml.SafeLoader)
self.path_to_image = os.path.join(
*[self.abs_path, self.config_camera["image_dir_name"]])
else:
print("'{}' do not have Image dirctory, ex) {}".format(
path_to_image, os.path.join(*[path_to_image, "Images/"])))
exit()
path_list = glob.glob(
os.path.join(*[
self.path_to_image, "*{}".format(
self.config_camera['image_style'])
]))
path_list = sorted(path_list,
key=lambda x: int(x.split("/")[-1].split(".")[0]))
output_dir_path = os.path.join(*[self.abs_path, self.output_dir_name])
os.makedirs(output_dir_path, exist_ok=True)
with open(
os.path.join(*[
output_dir_path,
self.path_to_analyze_config.split("/")[-1]
]), "w") as f:
yaml.dump(self.config, f)
return path_list
def choise_function(self, encode_image, api_kind):
if api_kind == "Detect":
return self.api.detect(image_base64=encode_image,
return_landmark=1,
return_attributes=",".join(
self.config["Attribute_Detect"]))
elif api_kind == "Dence":
return self.api.face.thousandlandmark(image_base64=encode_image,
return_landmark="all")
def detect(self, path_list, api_kind):
print("-" * 10 + api_kind + "-" * 10)
output_list = []
p_bar = ProgressBar(maxval=len(path_list))
for i, each_file in enumerate(path_list):
p_bar.update(i)
encode_img = self.img_to_bese64(each_file)
output = self.choise_function(encode_image=encode_img,
api_kind=api_kind)
output["image_path"] = each_file
output_list.append(output)
if (api_kind == "Dence") and (self.wait_time is not None):
import time
time.sleep(int(self.wait_time)) #実行制限の関係から，ここの値を調整する必要がある．
json_name, file_name = self.get_json_name(api_kind)
self.save_output(json_name, file_name, output_list)
def compere(self, path_list, target_file=None):
if target_file is None:
print("target_file is None, pleasse give any path to image file")
exit()
print("-" * 10 + "Get Compere" + "-" * 10)
p_bar = ProgressBar(maxval=len(path_list))
output_list = []
if target_file is None:
target_file = self.config["target_file"]
target_img = self.img_to_bese64(target_file)
for i, each_file in enumerate(path_list):
p_bar.update(i)
if each_file != target_file:
each_img = self.img_to_bese64(each_file)
output = self.api.compare(image_base64_1=target_img,
image_base64_2=each_img)
output["image_path"] = each_img
output_list.append(output)
json_name, file_name = self.get_json_name(
"Compere", name=target_file.split("/")[-1].split(".")[0])
self.save_output(json_name, file_name, output_list)
def get_json_name(self, api_kind, name=None, name_concat=False):
if name is None:
if not name_concat:
return os.path.join(
*[self.abs_path, self.output_dir_name, api_kind
]), self.output_json_name + ".json"
else:
return os.path.join(*[
self.abs_path, self.output_dir_name, api_kind,
self.output_json_name + ".json"
])
else:
if not name_concat:
return os.path.join(
*[self.abs_path, self.output_dir_name, api_kind
]), name + ".json"
else:
return os.path.join(*[
self.abs_path, self.output_dir_name, api_kind, name +
".json"
])
def save_output(self, path, file_name, output_list):
os.makedirs(path, exist_ok=True)
with open(os.path.join(*[path, file_name]), "w") as f:
json.dump(output_list, f)
def img_to_bese64(self, target_file):
with open(target_file, 'rb') as f:
data = f.read()
return base64.b64encode(data)
def __call__(self, target_file=None, path_to_image=None):
path_list = self.get_dataset(path_to_image=path_to_image)
if self.config["USE_API_Detect"]:
self.detect(path_list, api_kind="Detect")
Detect_api_json_to_csv(
path_to_json=self.get_json_name("Detect", name_concat=True))()
if self.config["USE_API_Compare"]:
self.compere(path_list, target_file)
if self.config["USE_API_Dense_Facial_Landmarks"]:
self.detect(path_list, "Dence")
Dence_api_json_to_csv(
path_to_json=self.get_json_name("Dence", name_concat=True))()
sys.exit()
