def face_search(request_id, b64_data, group_id='DEFAULT', max_user_num=5):
# 最多返回5个相似用户
max_user_num = min(5, max_user_num)
# 并行获取特征值 -- 分别检测人脸， 人脸角度可以分别调整
#all_encodings, face_locations = predict_parallel(get_features_thread_db, b64_data, group_id,
# request_id=request_id, classifier='api')
# 并行获取特征值 -- 只检测人脸一次，人脸角度要一起调整 (由ALGORITHM['evo']['p_angle']确定)
all_encodings, face_locations, face_array = verify.get_features_b64_parallel(b64_data, request_id=request_id)
if len(face_locations)==0: # 未取得特征值
return []
# 先使用knn分类器搜索（临时特征库）
predictions = predict_parallel(predict_thread_db, all_encodings, group_id,
request_id=request_id, classifier='knn', data_type='encodings')
# 如果未找到，再使用深度网络分类器（全量特征库）
if len(predictions)==0 or predictions[0][0]=='unknown':
print('search using keras classifier')
plus_encodings = [all_encodings['vgg']['None']+all_encodings['evo']['None']]
predictions = predict_K(plus_encodings, group_id,
model_path=TRAINED_MODEL_PATH,
face_algorithm="plus",
data_type="encodings")
# 如果仍未识别，返回空
if len(predictions)==0 or predictions[0][0]=='unknown':
return []
# 准备返回结果
user_list = []
for i in range(min(max_user_num, len(predictions))):
user_id, box, score, _ = predictions[i]
info = user_info(group_id, user_id)
user_list.append({
'user_id' : user_id,
'mobile_tail' : info['mobile'][:-4], # 手机后4位
'name' : info['name'], # 用户姓名
'location' : face_locations[0], # 只有一个人脸坐标
'score' : score,
})
# 只记录有结果的人脸数据，用于后面数据增强, 图片在预测时已保存
if len(user_list)>0:
face_save_to_temp(group_id, request_id, 'face_search', user_list, image=face_array[0])
#face_save_to_temp(group_id, request_id, 'face_search', user_list)
return user_list
sys.exit(2)
test_thing = sys.argv[1]
if os.path.isdir(test_thing):
images = os.listdir(test_thing)
images = [os.path.join(test_thing, i) for i in images]
else:
images = [test_thing]
# Using the trained classifier, make predictions for unknown images
for image_file in images:
print("Looking for faces in {}".format(image_file))
# Find all people in the image using a trained classifier model
# Note: You can pass in either a classifier file name or a classifier model instance
start_time = datetime.now()
predictions = predict_parallel(predict_thread, image_file)
print('[Time taken: {!s}]'.format(datetime.now() - start_time))
# Print results on the console
for name, (top, right, bottom, left), distance, count in predictions:
print("- Found {} at ({}, {}), distance={}, count={}".format(
name, left, top, distance, count))
if len(predictions) == 0:
print('Face not found!')
# Display results overlaid on an image
#knn.show_prediction_labels_on_image(image_file, predictions)
# Using the trained classifier, make predictions for unknown images
for image_file in images:
print("Looking for faces in {}".format(image_file))
with open(image_file, 'rb') as f:
image_data = f.read()
image_b64 = base64.b64encode(image_data)
# Find all people in the image using a trained classifier model
# Note: You can pass in either a classifier file name or a classifier model instance
#predictions = api_func.face_search('', image_b64, group_id)
#start_time = datetime.now()
predictions = predict_parallel(predict_thread_db, image_b64, group_id, classifier=classifier)
#print('[Time taken: {!s}]'.format(datetime.now() - start_time))
# Print results on the console
for name, (top, right, bottom, left), distance, count in predictions:
print("- Found {} at ({}, {}), distance={}, count={}".format(name, left, top, distance, count))
#for i in predictions:
# print("- Found {} at {}, distance={}".format(i['user_id'], i['location'], i['score']))
if len(predictions)==0:
print('Face not found!')
#print(predictions)
# Display results overlaid on an image
#knn.show_prediction_labels_on_image(image_file, predictions)
correct = 0
wrong = 0
fail = 0
multi = 0 # 匹配多个结果
second = 0 # 不是首个匹配结果
start_time = datetime.now()
for face in faces_list[-last_face_num:]: # 指定人的倒数几个人脸，用于使用train做测试
# 识别每个人脸
r = dbport.face_info(face)
if r is None:
continue
# 并行识别
predictions = predict_parallel(predict_thread_db,
r['encodings'],
group_id,
'encodings',
classifier=classifier)
# Print results on the console
if len(predictions) == 0:
fail += 1
else:
n = 0
bingo = 0
name_list = []
for name, (top, right, bottom,
left), distance, count in predictions:
if name == p:
if n == 0:
correct += 1