Python为人脸照片添加口罩实战

效果展示

数据集展示

数据集来源：使用了开源数据集FaceMask_CelebA

github地址：https://github.com/sevenHsu/FaceMask_CelebA.git

部分人脸数据集：

口罩样本数据集：

为人脸照片添加口罩代码

这部分有个库face_recognition需要安装，如果之前没有用过的小伙伴可能得费点功夫。
Face Recognition 库主要封装了dlib这一 C++ 图形库，通过 Python 语言将它封装为一个非常简单就可以实现人脸识别的 API 库，屏蔽了人脸识别的算法细节，大大降低了人脸识别功能的开发难度。

#!/usr/bin/env python
# -*- coding: utf-8 -*-
# @Author  : 2014Vee
import os
import numpy as np
from PIL import Image, ImageFile

__version__ = '0.3.0'

IMAGE_DIR = os.path.dirname('E:/play/FaceMask_CelebA-master/facemask_image/')
WHITE_IMAGE_PATH = os.path.join(IMAGE_DIR, 'front_14.png')
BLUE_IMAGE_PATH = os.path.join(IMAGE_DIR, 'front_14.png')
SAVE_PATH = os.path.dirname('E:/play/FaceMask_CelebA-master/save/synthesis/')
SAVE_PATH2 = os.path.dirname('E:/play/FaceMask_CelebA-master/save/masks/')

class FaceMasker:
   KEY_FACIAL_FEATURES = ('nose_bridge', 'chin')

def __init__(self, face_path, mask_path, white_mask_path, save_path, save_path2, model='hog'):
       self.face_path = face_path
       self.mask_path = mask_path
       self.save_path = save_path
       self.save_path2 = save_path2
       self.white_mask_path = white_mask_path
       self.model = model
       self._face_img: ImageFile = None
       self._black_face_img = None
       self._mask_img: ImageFile = None
       self._white_mask_img = None

def mask(self):
       import face_recognition

face_image_np = face_recognition.load_image_file(self.face_path)
       face_locations = face_recognition.face_locations(face_image_np, model=self.model)
       face_landmarks = face_recognition.face_landmarks(face_image_np, face_locations)
       self._face_img = Image.fromarray(face_image_np)
       self._mask_img = Image.open(self.mask_path)
       self._white_mask_img = Image.open(self.white_mask_path)
       self._black_face_img = Image.new('RGB', self._face_img.size, 0)

found_face = False
       for face_landmark in face_landmarks:
           # check whether facial features meet requirement
           skip = False
           for facial_feature in self.KEY_FACIAL_FEATURES:
               if facial_feature not in face_landmark:
                   skip = True
                   break
           if skip:
               continue

# mask face
           found_face = True
           self._mask_face(face_landmark)

if found_face:
           # save
           self._save()
       else:
           print('Found no face.')

def _mask_face(self, face_landmark: dict):
       nose_bridge = face_landmark['nose_bridge']
       nose_point = nose_bridge[len(nose_bridge) * 1 // 4]
       nose_v = np.array(nose_point)

chin = face_landmark['chin']
       chin_len = len(chin)
       chin_bottom_point = chin[chin_len // 2]
       chin_bottom_v = np.array(chin_bottom_point)
       chin_left_point = chin[chin_len // 8]
       chin_right_point = chin[chin_len * 7 // 8]

# split mask and resize
       width = self._mask_img.width
       height = self._mask_img.height
       width_ratio = 1.2
       new_height = int(np.linalg.norm(nose_v - chin_bottom_v))

# left
       mask_left_img = self._mask_img.crop((0, 0, width // 2, height))
       mask_left_width = self.get_distance_from_point_to_line(chin_left_point, nose_point, chin_bottom_point)
       mask_left_width = int(mask_left_width * width_ratio)
       mask_left_img = mask_left_img.resize((mask_left_width, new_height))

# right
       mask_right_img = self._mask_img.crop((width // 2, 0, width, height))
       mask_right_width = self.get_distance_from_point_to_line(chin_right_point, nose_point, chin_bottom_point)
       mask_right_width = int(mask_right_width * width_ratio)
       mask_right_img = mask_right_img.resize((mask_right_width, new_height))

# merge mask
       size = (mask_left_img.width + mask_right_img.width, new_height)
       mask_img = Image.new('RGBA', size)
       mask_img.paste(mask_left_img, (0, 0), mask_left_img)
       mask_img.paste(mask_right_img, (mask_left_img.width, 0), mask_right_img)

# rotate mask
       angle = np.arctan2(chin_bottom_point[1] - nose_point[1], chin_bottom_point[0] - nose_point[0])
       rotated_mask_img = mask_img.rotate(angle, expand=True)

# calculate mask location
       center_x = (nose_point[0] + chin_bottom_point[0]) // 2
       center_y = (nose_point[1] + chin_bottom_point[1]) // 2

offset = mask_img.width // 2 - mask_left_img.width
       radian = angle * np.pi / 180
       box_x = center_x + int(offset * np.cos(radian)) - rotated_mask_img.width // 2
       box_y = center_y + int(offset * np.sin(radian)) - rotated_mask_img.height // 2

# add mask
       self._face_img.paste(mask_img, (box_x, box_y), mask_img)

# split mask and resize
       width = self._white_mask_img.width
       height = self._white_mask_img.height
       width_ratio = 1.2
       new_height = int(np.linalg.norm(nose_v - chin_bottom_v))

# left
       mask_left_img = self._white_mask_img.crop((0, 0, width // 2, height))
       mask_left_width = self.get_distance_from_point_to_line(chin_left_point, nose_point, chin_bottom_point)
       mask_left_width = int(mask_left_width * width_ratio)
       mask_left_img = mask_left_img.resize((mask_left_width, new_height))

# right
       mask_right_img = self._white_mask_img.crop((width // 2, 0, width, height))
       mask_right_width = self.get_distance_from_point_to_line(chin_right_point, nose_point, chin_bottom_point)
       mask_right_width = int(mask_right_width * width_ratio)
       mask_right_img = mask_right_img.resize((mask_right_width, new_height))

# merge mask
       size = (mask_left_img.width + mask_right_img.width, new_height)
       mask_img = Image.new('RGBA', size)
       mask_img.paste(mask_left_img, (0, 0), mask_left_img)
       mask_img.paste(mask_right_img, (mask_left_img.width, 0), mask_right_img)

# rotate mask
       angle = np.arctan2(chin_bottom_point[1] - nose_point[1], chin_bottom_point[0] - nose_point[0])
       rotated_mask_img = mask_img.rotate(angle, expand=True)

# calculate mask location
       center_x = (nose_point[0] + chin_bottom_point[0]) // 2
       center_y = (nose_point[1] + chin_bottom_point[1]) // 2

offset = mask_img.width // 2 - mask_left_img.width
       radian = angle * np.pi / 180
       box_x = center_x + int(offset * np.cos(radian)) - rotated_mask_img.width // 2
       box_y = center_y + int(offset * np.sin(radian)) - rotated_mask_img.height // 2

# add mask
       self._black_face_img.paste(mask_img, (box_x, box_y), mask_img)

def _save(self):
       path_splits = os.path.splitext(self.face_path)
       # new_face_path = self.save_path + '/' + os.path.basename(self.face_path) + '-with-mask' + path_splits[1]
       # new_face_path2 = self.save_path2 + '/' + os.path.basename(self.face_path) + '-binary' + path_splits[1]
       new_face_path = self.save_path + '/' + os.path.basename(self.face_path) + '-with-mask' + path_splits[1]
       new_face_path2 = self.save_path2 + '/'  + os.path.basename(self.face_path) + '-binary' + path_splits[1]
       self._face_img.save(new_face_path)
       self._black_face_img.save(new_face_path2)

#         print(f'Save to {new_face_path}')

@staticmethod
   def get_distance_from_point_to_line(point, line_point1, line_point2):
       distance = np.abs((line_point2[1] - line_point1[1]) * point[0] +
                         (line_point1[0] - line_point2[0]) * point[1] +
                         (line_point2[0] - line_point1[0]) * line_point1[1] +
                         (line_point1[1] - line_point2[1]) * line_point1[0]) / \
                  np.sqrt((line_point2[1] - line_point1[1]) * (line_point2[1] - line_point1[1]) +
                          (line_point1[0] - line_point2[0]) * (line_point1[0] - line_point2[0]))
       return int(distance)

# FaceMasker("/home/aistudio/data/人脸.png", WHITE_IMAGE_PATH, True, 'hog').mask()

from pathlib import Path

images = Path("E:/play/FaceMask_CelebA-master/bbox_align_celeba").glob("*")
cnt = 0
for image in images:
   if cnt < 1:
       cnt += 1
       continue
   FaceMasker(image, BLUE_IMAGE_PATH, WHITE_IMAGE_PATH, SAVE_PATH, SAVE_PATH2, 'hog').mask()
   cnt += 1
   print(f"正在处理第{cnt}张图片，还有{99 - cnt}张图片")

掩膜生成代码

这部分其实就是对使用的口罩样本的二值化，因为后续要相关模型会用到

import os
from PIL import Image

# 源目录
# MyPath = 'E:/play/FaceMask_CelebA-master/facemask_image/'
MyPath = 'E:/play/FaceMask_CelebA-master/save/masks/'
# 输出目录
OutPath = 'E:/play/FaceMask_CelebA-master/save/Binarization/'

def processImage(filesoure, destsoure, name, imgtype):
   '''
   filesoure是存放待转换图片的目录
   destsoure是存在输出转换后图片的目录
   name是文件名
   imgtype是文件类型
   '''
   imgtype = 'bmp' if imgtype == '.bmp' else 'png'
   # 打开图片
   im = Image.open(filesoure + name)
   # =============================================================================
   #     #缩放比例
   #     rate =max(im.size[0]/640.0 if im.size[0] > 60 else 0, im.size[1]/1136.0 if im.size[1] > 1136 else 0)
   #     if rate:
   #         im.thumbnail((im.size[0]/rate, im.size[1]/rate))
   # =============================================================================

img = im.convert("RGBA")
   pixdata = img.load()
   # 二值化
   for y in range(img.size[1]):
       for x in range(img.size[0]):
           if pixdata[x, y][0] < 90:
               pixdata[x, y] = (0, 0, 0, 255)

for y in range(img.size[1]):
       for x in range(img.size[0]):
           if pixdata[x, y][1] < 136:
               pixdata[x, y] = (0, 0, 0, 255)

for y in range(img.size[1]):
       for x in range(img.size[0]):
           if pixdata[x, y][2] > 0:
               pixdata[x, y] = (255, 255, 255, 255)
   img.save(destsoure + name, imgtype)

def run():
   # 切换到源目录，遍历源目录下所有图片
   os.chdir(MyPath)
   for i in os.listdir(os.getcwd()):
       # 检查后缀
       postfix = os.path.splitext(i)[1]
       name = os.path.splitext(i)[0]
       name2 = name.split('.')
       if name2[1] == 'jpg-binary' or name2[1] == 'png-binary':
           processImage(MyPath, OutPath, i, postfix)

if __name__ == '__main__':
   run()

来源：https://blog.csdn.net/qq_52118067/article/details/124334872