def main():
my = php.kit()
# 人臉68特徵點模型路徑
predictor_path = "shape_predictor_68_face_landmarks.dat"
# 人臉辨識模型路徑
face_rec_model_path = "dlib_face_recognition_resnet_model_v1.dat"
# 載入人臉檢測器
detector = dlib.get_frontal_face_detector()
# 載入人臉特徵點檢測器
sp = dlib.shape_predictor(predictor_path)
# 載入人臉辨識檢測器
facerec = dlib.face_recognition_model_v1(face_rec_model_path)
fp = my.glob("rec\\pic\\*.jpg")
np_dir = "rec\\numpy"
# 針對比對資料夾裡每張圖片做比對:
# 1.人臉偵測
# 2.特徵點偵測
# 3.取得描述子
descriptors = []
step = 0
totals = len(fp)
for f in fp:
print("Run ... " + str(step) + " / " + str((totals - 1)))
step = step + 1
base = os.path.basename(f)
mn = my.mainname(f)
op_txt = np_dir + "\\" + mn
if my.is_file(op_txt + ".npy") == True:
continue
img = io.imread(f)
# 1.人臉偵測
dets = detector(img, 1)
for k, d in enumerate(dets):
# 2.特徵點偵測
shape = sp(img, d)
# 3.取得描述子，128維特徵向量
face_descriptor = facerec.compute_face_descriptor(img, shape)
# 轉換numpy array格式
v = numpy.array(face_descriptor)
#descriptors.append(v)
#print(descriptors)
# https://ithelp.ithome.com.tw/articles/10196167
numpy.save(op_txt, v)
# -*- coding: UTF-8 -*-
import sys, os, dlib, glob, numpy
from skimage import io
import cv2
import imutils
import php
my = php.kit()
if len(sys.argv) != 2:
print("缺少要辨識的圖片名稱")
exit()
# 人臉68特徵點模型路徑
predictor_path = "shape_predictor_68_face_landmarks.dat"
# 人臉辨識模型路徑
face_rec_model_path = "dlib_face_recognition_resnet_model_v1.dat"
# 比對人臉圖片資料夾名稱
faces_folder_path = "./rec/pic"
# 比對人臉解算資料夾名稱
faces_numpy_folder_path = "./rec/numpy"
# 需要辨識的人臉圖片名稱
img_path = sys.argv[1]
# 載入人臉檢測器
detector = dlib.get_frontal_face_detector()
def camLoop():
my = php.kit()
# 比對人臉解算資料夾名稱
faces_numpy_folder_path = "./rec/numpy"
# 人臉辨識模型路徑
face_rec_model_path = "dlib_face_recognition_resnet_model_v1.dat"
# 載入人臉辨識檢測器
facerec = dlib.face_recognition_model_v1(face_rec_model_path)
# 比對人臉描述子列表
descriptors = []
# 比對人臉名稱列表
candidate = []
# 針對比對資料夾裡每張圖片做比對:
# 1.人臉偵測
# 2.特徵點偵測
# 3.取得描述子
for f in my.glob(faces_numpy_folder_path+"\\*.npy"):
base = os.path.basename(f)
# 依序取得圖片檔案人名
candidate.append(os.path.splitext(base)[ 0])
# from : https://ithelp.ithome.com.tw/articles/10196167
v = numpy.load(f)
descriptors.append(v)
#選擇第一隻攝影機
cap = cv2.VideoCapture(0, cv2.CAP_DSHOW)
#調整預設影像大小，預設值很大，很吃效能
cap.set(cv2. CAP_PROP_FRAME_WIDTH, 320)
cap.set(cv2. CAP_PROP_FRAME_HEIGHT, 240)
#取得預設的臉部偵測器
detector = dlib.get_frontal_face_detector()
#根據shape_predictor方法載入68個特徵點模型，此方法為人臉表情識別的偵測器
predictor = dlib.shape_predictor( 'shape_predictor_68_face_landmarks.dat')
#當攝影機打開時，對每個frame進行偵測
step_while = 0
x1=0
last_rec_name=""
img = cv2.imread('qrcode.png')
use_code = ['enter']
while(cap.isOpened()):
#time.sleep(0.5)
#讀出frame資訊
ret, frame = cap.read()
step_while=step_while+1
#取出偵測的結果
#print(step_while)
if step_while>120:
x1=0
last_rec_name = ""
step_while = 0
if step_while % 60==0:
for code in decode(frame):
if code.data.decode('utf-8') in use_code:
year=datetime.datetime.now().strftime('%Y')
month=datetime.datetime.now().strftime('%m')
day=datetime.datetime.now().strftime('%d')
current_time=datetime.datetime.now().strftime('%H:%M:%S')
tsv_writer.write_tsv("VIP",current_time,year,month,day)
print('You can enter!')
time.sleep(3)
dist = []
#偵測人臉
face_rects, scores, idx = detector.run(frame, 0)
for i, d in enumerate(face_rects):
x1 = d.left()
y1 = d.top()
x2 = d.right()
y2 = d.bottom()
#text = " %2.2f ( %d )" % (scores[i], idx[i])
#繪製出偵測人臉的矩形範圍
cv2.rectangle(frame, (x1, y1), (x2, y2), ( 0, 255, 0), 4, cv2. LINE_AA)
#標上人臉偵測分數與人臉方向子偵測器編號
#cv2.putText(frame, text, (x1, y1), cv2. FONT_HERSHEY_DUPLEX,
#0.7, ( 255, 255, 255), 1, cv2. LINE_AA)
#給68特徵點辨識取得一個轉換顏色的frame
landmarks_frame = cv2.cvtColor(frame, cv2. COLOR_BGR2RGB)
#找出特徵點位置
shape = predictor(landmarks_frame, d)
face_descriptor = facerec.compute_face_descriptor(frame, shape)
d_test = numpy.array(face_descriptor)
# 計算歐式距離
for j in descriptors:
dist_ = numpy.linalg.norm(j - d_test)
dist.append(dist_)
#繪製68個特徵點
#for i in range( 68):
# cv2.circle(frame,(shape.part(i).x,shape.part(i).y), 3,( 0, 0, 255), 2)
# cv2.putText(frame, str(i),(shape.part(i).x,shape.part(i).y),cv2. FONT_HERSHEY_COMPLEX, 0.5,( 255, 0, 0), 1)
#輸出到畫面
# 將比對人名和比對出來的歐式距離組成一個dict
c_d = dict( zip(candidate,dist))
# 根據歐式距離由小到大排序
#cd_sorted = sorted(c_d.iteritems(), key = lambda d:d[ 1])
cd_sorted = sorted(c_d.items(), key=lambda kv: kv[1])
# 取得最短距離就為辨識出的人名
#print(cd_sorted)
if cd_sorted[0][1]<0.4:
rec_name = cd_sorted[0][0]
m = my.explode("#",rec_name)
last_rec_name = m[0]
#首字大寫
last_rec_name = my.strtolower(last_rec_name)
last_rec_name = last_rec_name.capitalize()
step_while = 0
# 將辨識出的人名印到圖片上面
cv2.putText(frame, last_rec_name, (x1, y1), cv2. FONT_HERSHEY_SIMPLEX , 1, ( 255, 255, 255), 2, cv2. LINE_AA)
year=datetime.datetime.now().strftime('%Y')
month=datetime.datetime.now().strftime('%m')
day=datetime.datetime.now().strftime('%d')
current_time=datetime.datetime.now().strftime('%H:%M:%S')
print(year," ",month," ",day," ",current_time,"員工ID:",last_rec_name)
winsound.MessageBeep()
tsv_writer.write_tsv(last_rec_name,current_time,year,month,day)
break
#print(m[0])
if x1!=0:
cv2.rectangle(frame, (x1, y1), (x2, y2), ( 0, 255, 0), 4, cv2. LINE_AA)
if last_rec_name!="":
cv2.putText(frame, last_rec_name, (x1, y1), cv2. FONT_HERSHEY_SIMPLEX , 1, ( 255, 255, 255), 2, cv2. LINE_AA)
cv2.imshow( "Face Detection, press esc to quit", frame)
#如果按下ESC键，就退出
if cv2.waitKey( 10) == 27:
break
#釋放記憶體
cap.release()
#關閉所有視窗
cv2.destroyAllWindows()