Windows 10+dlib19.17+face_recognition：使用人脸识别对家庭照片进行分类，4-6秒/张（一）单线程

闪闪·Style

3148人浏览 · 2019-08-09 09:37:31

闪闪·Style · 2019-08-09 09:37:31 发布

在Windows 10环境中安装好了dlib19.17和face_recognition，具体过程请参考：

https://blog.csdn.net/weixin_41943311/article/details/91866987

https://blog.csdn.net/weixin_41943311/article/details/98482615

想一想能干啥用？第一个冒出来的想法是用人脸识别对存在电脑中的家庭照片（常规大小，比如3~5MB/张）进行分类，简单说：就是先识别某人的人脸，然后对目录下的所有照片进行遍历，找到有某人的所有照片，并把这些照片统一复制到一个新的目录下。

简单写了一个程序，代码如下（可以支持中文目录和中文文件名）：

# -*- coding: UTF-8 -*-

import dlib
import face_recognition
import numpy as np
from datetime import datetime
import os
import shutil

# 从图片中加载已知的人脸并获得编码
steve_image = face_recognition.load_image_file("f:/images/steve.jpg")
steve_face_encoding = face_recognition.face_encodings(steve_image)[0]

lucy_image = face_recognition.load_image_file("f:/images/lucy.jpg")
lucy_face_encoding = face_recognition.face_encodings(lucy_image)[0]

known_face_encodings = [
    steve_face_encoding,
    lucy_face_encoding,
]

known_face_names = [
    "Steve",
    "Lucy",
]

# 测试起始时间
t1 = datetime.now()
t10 = t1 - t1
t20 = t10
t30 = t10
t40 = t10
t50 = t10
count_checked, count_copied = 0, 0

# 遍历目录下的所有.jpg文件
f = os.walk("f:\images")
for path,d,filelist in f:
    for filename in filelist:
        if filename.endswith('jpg'):
            image_path = os.path.join(path, filename)
            # 加载图片
            t00 = datetime.now()
            unknown_image = face_recognition.load_image_file(image_path)
            t10 += datetime.now() - t00
            count_checked += 1

            # 找到图中所有人脸的位置
            t00 = datetime.now()
            face_locations = face_recognition.face_locations(unknown_image)
            #face_locations = face_recognition.face_locations(unknown_image, number_of_times_to_upsample=0, model="cnn")
            t20 += datetime.now() - t00

            # 根据位置加载人脸编码的列表
            t00 = datetime.now()
            face_encodings = face_recognition.face_encodings(unknown_image, face_locations)
            t30 += datetime.now() - t00

            # 遍历所有人脸编码，与已知人脸对比
            for (top, right, bottom, left), face_encoding in zip(face_locations, face_encodings):
                # 获得对比结果，tolerance值越低比对越严格
                t00 = datetime.now()
                matches = face_recognition.compare_faces(known_face_encodings, face_encoding, tolerance=0.4)
                t40 += datetime.now() - t00

                # 获得比对成功的姓名（未做进一步处理），并复制文件到指定目录
                name = "Unknown"
                if True in matches:
                    first_match_index = matches.index(True)
                    name = known_face_names[first_match_index]
                    # 若同一个图片中有多个已知人脸，重复复制文件会报错
                    try:
                        t00 = datetime.now()
                        shutil.copy(image_path,"f:/images_family")
                        t50 += datetime.now() - t00
                        count_copied += 1
                        break
                    except shutil.Error:
                        break

# 测试结束时间
t2 = datetime.now()
# 显示总的时间开销
print('%d pictures checked, and %d pictures copied with known faces.' %(count_checked, count_copied))
print('time spend: %d seconds, %d microseconds.' %((t2-t1).seconds, (t2-t1).microseconds))
print('load_image_file time: %d seconds, %d microseconds.' %(t10.seconds, t10.microseconds))
print('face_locations  time: %d seconds, %d microseconds.' %(t20.seconds, t20.microseconds))
print('face_encodings  time: %d seconds, %d microseconds.' %(t30.seconds, t30.microseconds))
print('compare_faces   time: %d seconds, %d microseconds.' %(t40.seconds, t40.microseconds))
print('shutil.copy     time: %d seconds, %d microseconds.' %(t50.seconds, t50.microseconds))

实际运行的目录（f:\images）下有9个文件夹，102个文件，其中有99张照片，共311MB，平均每张照片大小在3MB左右。程序在PyCharm中运行，运行时CPU负载为16%-33%之间（没有用到GPU，因为使用800x600以上图片识别人脸时，NVIDIA GeForce GTX 1060动态内存分配时内存不足报错）：