经过前两篇入门级的介绍,不知道各位和我一样的小白,感受到py做人脸识别的乐趣没有?上一篇中的识别效果满意吗?答案是不满意啊!
通过费力搭建开发环境,发现识别效果差强人意。怎么办?
我们可以通过GITHUB来寻找相关的实验项目。
下面上一个可以调节识别阈值的代码段,大家看看会出现什么!!!
# -*- coding: utf-8 -*-
# 摄像头头像识别
import face_recognition
import cv2
from os import listdir
source = "×××××" # 摄像头的rtsp地址
cam = cv2.VideoCapture(0)
filepath = 'C:/Users/Administrator\Desktop/recognitiongface1/face_photos' # 已知人脸图片文件夹 注意 如果会员图片后缀不是jpg 需要进行修改
filename_list = listdir(filepath)
known_face_names = []
known_face_encodings = []
a = 0
for filename in filename_list: # 依次读入列表中的内容
a += 1
if filename.endswith('jpg'): # 后缀名'jpg'匹对
known_face_names.append(filename[:-4]) # 把文件名字的后四位.jpg去掉获取人名
file_str = filepath + '/' + filename
a_images = face_recognition.load_image_file(file_str)
# print(file_str)
a_face_encoding = face_recognition.face_encodings(a_images)[0]
known_face_encodings.append(a_face_encoding)
print(known_face_names, a)
face_locations = []
face_encodings = []
face_names = []
process_this_frame = True
while (cam.isOpened()):
# 读取摄像头画面
ret, frame = cam.read()
if not ret:
# 等同于 if ret is not none
break
# 改变摄像头图像的大小,图像小,所做的计算就少
small_frame = cv2.resize(frame, (0, 0), fx=0.33, fy=0.33)
# opencv的图像是BGR格式的,而我们需要是的RGB格式的,因此需要进行一个转换。
rgb_small_frame = small_frame[:, :, ::-1]
# Only process every other frame of video to save time
if process_this_frame:
# 根据encoding来判断是不是同一个人,是就输出true,不是为flase
face_locations = face_recognition.face_locations(rgb_small_frame)
face_encodings = face_recognition.face_encodings(rgb_small_frame, face_locations)
face_names = []
for face_encoding in face_encodings:
# 默认为unknown
matches = face_recognition.compare_faces(known_face_encodings, face_encoding, tolerance=0.38)
# 阈值太低容易造成无法成功识别人脸,太高容易造成人脸识别混淆 默认阈值tolerance为0.6
# print(matches)
name = "Unknown"
# if match[0]:
# name = "michong"
# If a match was found in known_face_encodings, just use the first one.
if True in matches:
first_match_index = matches.index(True)
name = known_face_names[first_match_index]
face_names.append(name)
process_this_frame = not process_this_frame
# 将捕捉到的人脸显示出来
for (top, right, bottom, left), name in zip(face_locations, face_names):
# Scale back up face locations since the frame we detected in was scaled to 1/4 size
# 由于我们检测到的帧被缩放到1/4大小,所以要缩小面位置
top *= 3
right *= 3
bottom *= 3
left *= 3
# 矩形框
cv2.rectangle(frame, (left, top), (right, bottom), (0, 0, 255), 2)
# 引入ft2中的字体
# 加上标签
cv2.rectangle(frame, (left, bottom - 20), (right, bottom), (0, 0, 255), cv2.FILLED)
font = cv2.FONT_HERSHEY_DUPLEX
cv2.putText(frame, name, (left + 6, bottom - 6), font, 0.8, (255, 255, 255), 1)
# frame = ft.draw_text(frame,(left + 6, bottom - 6), name, 1.0, (255, 255, 255))
# def draw_text(self, image, pos, text, text_size, text_color)
# Display
cv2.imshow('monitor', frame)
if cv2.waitKey(1) & 0xFF == 27:
break
cam.release()
cv2.destroyAllWindows()
这样之后呢?注意阅读代码注释,修改:
matches =face_recognition.compare_faces(known_face_encodings, face_encoding, tolerance=0.38)
是不是识别率发生变化了?注意、阈值设定太低的话,识别不出人脸哟。阈值太高的话、识别准确率很低的。
1、掏出你82年的手机,自拍一张大头贴,对准摄像头,看看能识别出来不?
2、上述方式也能识别的话,是不是有些儿戏,有些不严谨,有漏洞。。。
3、怎么办呢?预知详情,请听下回分解!