基于face_recognition库的人脸识别、对比实现

github

介绍

本项目是世界上最简洁的人脸识别库，你可以使用Python和命令行工具提取、识别、操作人脸。

本项目的人脸识别是基于业内领先的C++开源库 dlib中的深度学习模型，用Labeled Faces in the Wild人脸数据集进行测试，有高达99.38%的准确率。但对小孩和亚洲人脸的识别准确率尚待提升。

Labeled Faces in the Wild是美国麻省大学安姆斯特分校（University of Massachusetts Amherst)制作的人脸数据集，该数据集包含了从网络收集的13,000多张面部图像。

本项目提供了简易的face_recognition命令行工具，你可以用它处理整个文件夹里的图片。

实现

#!/usr/bin/env python
# -*- coding: utf-8 -*-
'''
@File    :   main.py
@Time    :   2023/09/25 15:30:22
@Author  :   Carry
@Version :   1.0
@Desc    :   人脸识别、人脸比对
'''

import time
import cv2
import sys
import face_recognition
import gradio as gr


def draw_face_rectangles(image, locations):
    """
    # 定义一个函数来绘制人脸区域
    """
    for location in locations:
        top, right, bottom, left = location
        cv2.rectangle(image, (left, top), (right, bottom), (0, 255, 0), 2)


def find_largest_face(face_locations):
    """
    # 找到每张图片中最大的人脸
    """
    if not face_locations:
        return None
    largest_face = face_locations[0]
    largest_area = (largest_face[2] - largest_face[0]) * (largest_face[1] - largest_face[3])
    for face_location in face_locations:
        area = (face_location[2] - face_location[0]) * (face_location[1] - face_location[3])
        if area > largest_area:
            largest_area = area
            largest_face = face_location
    return largest_face


def is_same_person(image1, image2, local=False, save_img=False):
    """
    # 判断两张图片中的人脸是否是同一个人
    """
    if local:
        # 加载两张图片
        image1 = face_recognition.load_image_file(image1)
        image2 = face_recognition.load_image_file(image2)

    # 在每张图片中找到所有人脸的面部位置
    t1 = time.time()
    face_locations1 = face_recognition.face_locations(image1)
    face_locations2 = face_recognition.face_locations(image2)
    t2 = time.time()
    print("face location time:", t2-t1)

    if save_img:
        # 绘制人脸区域并保存图像
        image1_with_faces = image1.copy()
        draw_face_rectangles(image1_with_faces, face_locations1)
        cv2.imwrite("image1_with_faces.jpg", image1_with_faces)

        image2_with_faces = image2.copy()
        draw_face_rectangles(image2_with_faces, face_locations2)
        cv2.imwrite("image2_with_faces.jpg", image2_with_faces)

    # 找到每张图片中最大的人脸
    largest_face1 = find_largest_face(face_locations1)
    largest_face2 = find_largest_face(face_locations2)
    print(f"largest face1: {largest_face1}")
    print(f"largest face2: {largest_face2}")

    if largest_face1 is None or largest_face2 is None:
        return "未能找到图片中的人脸"
    else:
        # 提取人脸特征
        t1 = time.time()
        face_encodings1 = face_recognition.face_encodings(image1, [largest_face1], num_jitters=1, model="large")[0]
        face_encodings2 = face_recognition.face_encodings(image2, [largest_face2], num_jitters=1, model="large")[0]
        t2 = time.time()
        print("face encoding time:", t2-t1)

        face_distances = face_recognition.face_distance([face_encodings1], face_encodings2)
        print(f"face distance: {face_distances}")

        results = face_recognition.compare_faces([face_encodings1], face_encodings2, tolerance=0.4)
        print(f"compare faces: {results}")

        if results[0]:
            return f"这两张图片中的人脸是同一个人。score(越趋近0越相似,当前阈值0.4):{face_distances[0]}"
        else:
            return f"这两张图片中的人脸不是同一个人。score(越趋近0越相似,当前阈值0.4):{face_distances[0]}"


# 接收命令行输入的图片路径
# pic_1 = sys.argv[1]
# pic_2 = sys.argv[2]
# t1 = time.time()
# is_same_person(pic_1, pic_2, local=True)
# t2 = time.time()
# print("total time:", t2-t1)

# 创建 Gradio 交互式界面
iface = gr.Interface(fn=is_same_person, inputs=["image", "image"], outputs="text")
iface.launch(server_name="0.0.0.0")

运行

python main.py
http://localhost:7861/