def upload():
if request.method == 'POST':
f = request.files['file']
cont = f.read()
buf = np.frombuffer(cont, dtype=np.byte)
# print('buf:size is {}:{} '.format(buf, np.shape(buf)))
print('buf {} and buf size {}'.format(buf, np.shape(buf)))
img = np.expand_dims(cv2.imdecode(buf, cv2.IMREAD_COLOR), axis=0)
print('input image shape ', np.shape(img))
emb = faceNet_serving_V0.img_to_emb_feature(img, FACENET_CHANNEL)
print('emb size ', len(emb))
ret = {"file": f.filename, "emb": list(emb)}
# print('return: ', ret)
return json.dumps(ret)
def upload():
if request.method == 'POST':
try:
f = request.files['file'] # 301 file error
cont = f.read()
buf = np.frombuffer(cont, dtype=np.byte)
img = cv2.imdecode(buf, cv2.IMREAD_COLOR)
# print('image size', np.shape(img))
faces, _, det_arr = load_and_align_data(img)
print('face shape', np.shape(faces))
print('bounding boxes', det_arr)
if faces is None:
print('No Faces in this image!')
ret = {
"file": f.filename,
"code": 200,
"message": "Has_no_faces",
"result": "合规"
}
else:
det_arr_ser = np.reshape(det_arr, (1, np.size(det_arr)))
emb = faceNet_serving_V0.img_to_emb_feature(faces, FACENET_CHANNEL)
print('embedding shape ', np.shape(emb))
emb = list(emb)
num_face = int(len(emb)/128)
ret_not_pass = {}
data = []
for i in range(num_face):
emb_face = emb[i*128: (1+i)*128]
likely = cal_sim_new(emb_face, emb_dict)
theta = {'xijinping': [0.75],
'hujintao': [0.73],
'jiangzemin': [0.72],
'dengxiaoping': [0.60],
'wenjiabao': [0.71],
'maozedong': [0.85],
'zhouenlai': [0.78]}
diff_name = list([i for i, _ in theta.items()])
value = []
for name in diff_name:
theta[name].append(likely[name])
value.append(theta[name])
value = np.squeeze(value)
# print('dictionary theta', theta)
Flag_all = listsbigger(value[:,0], value[:,1])
if all(Flag_all):
value_max = np.max(value[:,1])
value_max_indice = np.argmax(value[:,1])
value_max_name = diff_name[value_max_indice]
ret_pass = {
"file": f.filename,
"code": 300,
"message": "Has_face_pass",
"result": "合规",
"user_name": str(value_max_name),
"score": str(value_max),
"det_arr": det_arr_ser.tolist()
}
else:
index_False = [i for i, x in enumerate(Flag_all) if not x]
max_term_tmp = value[index_False, :]
delta_score = max_term_tmp[:, 1] - max_term_tmp[:, 0]
delta_score_max_indice = np.argmax(delta_score)
max_term = max_term_tmp[delta_score_max_indice, 1]
max_name_tmp = [diff_name[i] for i in index_False]
max_name = max_name_tmp[delta_score_max_indice]
# print('max_name {} and its value {}'.format(max_name, max_term))
det_arr_tmp = np.array(det_arr)[i, :]
det_arr_ser = np.reshape(det_arr_tmp, (1, np.size(det_arr_tmp)))
data.append({
"face_id": str(i),
"user_name": str(max_name),
"score": str(max_term),
"det_arr": det_arr_ser.tolist()
})
ret_not_pass = {
"file": f.filename,
"code": 301,
"message": "敏感人物",
"result": "不合规",
"data": data,
}
if not bool(ret_not_pass):
ret = ret_pass
else:
ret = ret_not_pass
print(ret)
except Exception as e:
print('#cv encoder error!')
ret = {"file": f.filename,
"code": 301,
"message": "encoder_failed",
"result": "PASS"}
return json.dumps(ret)
def main():
# img = cv2.imdecode(buf, cv2.IMREAD_COLOR)
img = cv2.imread('F:/peoples_baidu/xijinping_baidu/10.jpg', cv2.IMREAD_COLOR)
faces, det_arr, bb_int = load_and_align_data(img)
for i in range(np.shape(faces)[0]):
# det_arr_plt = np.zeros((1,4), dtype=np.int32)
det_arr_plt = det_arr[i].astype(np.int32)
bb_int = bb_int[i]
cv2.imshow('img', img)
cv2.rectangle(img, (det_arr_plt[0],det_arr_plt[1]), (det_arr_plt[2],det_arr_plt[3]), (0,255,0))
cv2.rectangle(img, (bb_int[0], bb_int[1]),(bb_int[2],bb_int[3]), (255, 0, 255))
cv2.waitKey()
cv2.imshow('face', faces[i, :, : ,:])
cv2.waitKey()
if faces is None:
print('No Faces in this image!')
ret = {
"code": 200,
"message": "Has_no_faces",
"result": "合规",
}
else:
det_arr_ser = np.reshape(det_arr, (1, np.size(det_arr)))
print('input face shape ', np.shape(faces))
emb = faceNet_serving_V0.img_to_emb_feature(faces, FACENET_CHANNEL)
print('emb size', len(emb))
emb = list(emb)
num_face = int(len(emb)/512)
ret = {}
maximum = []
maximum_name = ['test']
for i in range(num_face):
emb_face = emb[i*512: (1+i)*512]
likely = cal_sim_new(emb_face, data_ave)
print('Likely ', likely)
maximum_name.append(max(likely, key=likely.get))
maximum.append(likely[max(likely, key=likely.get)])
th = 0.40
if max(maximum) < th:
ret = {
"code": 300,
"message": "Has_face_pass",
"result": "合规",
"det_arr": det_arr_ser.tolist(),
}
else:
index = list(np.where(np.array(maximum) > th)[0])
det_arr_tmp = np.array(det_arr)[index, :]
det_arr_ser = np.reshape(det_arr_tmp, (1, np.size(det_arr_tmp)))
data = []
for ii in index:
data.append(
{
"face_id": str(ii),
"user_name": maximum_name[ii+1],
"score": maximum[ii]
})
ret = {
"code": 301,
"message": "敏感人物",
"result": "不合规",
"data": data,
"det_arr": det_arr_ser.tolist()
}
return json.dumps(ret)
def upload():
if request.method == 'POST':
f = request.files['file'] # 301 file error
cont = f.read()
buf = np.frombuffer(cont, dtype=np.byte)
img = cv2.imdecode(buf, cv2.IMREAD_COLOR)
faces, det_arr = load_and_align_data(img)
print('file {} have #{} faces'.format(f, len(det_arr)))
if faces is None:
print('No Faces in this image!')
ret = {
"file": f.filename,
"code": 200,
"message": "Has_no_faces",
"result": "合规"
}
else:
det_arr_ser = np.reshape(det_arr, (1, np.size(det_arr)))
emb = faceNet_serving_V0.img_to_emb_feature(faces, FACENET_CHANNEL)
emb = list(emb)
num_face = int(len(emb) / 128)
ret = {}
maximum = []
maximum_name = ['test']
for i in range(num_face):
emb_face = emb[i * 128:(1 + i) * 128]
likely = cal_sim_new(emb_face, emb_data)
maximum_name.append(max(likely, key=likely.get))
maximum.append(likely[max(likely, key=likely.get)])
th = 0.40
if max(maximum) < th:
ret = {
"file": f.filename,
"code": 300,
"message": "Has_face_pass",
"result": "合规",
"det_arr": det_arr_ser.tolist()
}
else:
index = list(np.where(np.array(maximum) > th)[0])
det_arr_tmp = np.array(det_arr)[index, :]
det_arr_ser = np.reshape(det_arr_tmp,
(1, np.size(det_arr_tmp)))
data = []
for ii in index:
data.append({
"face_id": str(ii),
"user_name": maximum_name[ii + 1],
"score": maximum[ii]
})
ret = {
"file": f.filename,
"code": 301,
"message": "敏感人物",
"result": "不合规",
"data": data,
"det_arr": det_arr_ser.tolist()
}
return json.dumps(ret)