opencv&mediapipe 人脸检测+摄像头实时

Posted 2022-01-22 墙缝里的草

文章目录

• • 单张人脸关键点检测
  • 单张图像人脸检测
  • 摄像头实时关键点检测

单张人脸关键点检测

定义可视化图像函数
导入三维人脸关键点检测模型
导入可视化函数和可视化样式
读取图像
将图像模型输入，获取预测结果
BGR转RGB
将RGB图像输入模型，获取预测结果
预测人人脸个数
可视化人脸关键点检测效果
绘制人来脸和重点区域轮廓线，返回annotated_image
绘制人脸轮廓、眼睫毛、眼眶、嘴唇
在三维坐标中分别可视化人脸网格、轮廓、瞳孔

import cv2 as cv
import  mediapipe as mp
from tqdm import tqdm
import time
import  matplotlib.pyplot as plt

# 定义可视化图像函数
def look_img(img):
    img_RGB=cv.cvtColor(img,cv.COLOR_BGR2RGB)
    plt.imshow(img_RGB)
    plt.show()

# 导入三维人脸关键点检测模型
mp_face_mesh=mp.solutions.face_mesh
# help(mp_face_mesh.FaceMesh)

model=mp_face_mesh.FaceMesh(
    static_image_mode=True,#TRUE:静态图片/False:摄像头实时读取
    refine_landmarks=True,#使用Attention Mesh模型
    min_detection_confidence=0.5, #置信度阈值，越接近1越准
    min_tracking_confidence=0.5,#追踪阈值
)


# 导入可视化函数和可视化样式
mp_drawing=mp.solutions.drawing_utils
mp_drawing_styles=mp.solutions.drawing_styles

# 读取图像

img=cv.imread('img.png')

# look_img(img)

# 将图像模型输入，获取预测结果

# BGR转RGB
img_RGB=cv.cvtColor(img,cv.COLOR_BGR2RGB)

# 将RGB图像输入模型，获取预测结果

results=model.process(img_RGB)
# 预测人人脸个数
len(results.multi_face_landmarks)

print(len(results.multi_face_landmarks))

# 结果：1


# 可视化人脸关键点检测效果

# 绘制人来脸和重点区域轮廓线，返回annotated_image
annotated_image=img.copy()
if results.multi_face_landmarks: #如果检测出人脸
    for face_landmarks in results.multi_face_landmarks:#遍历每一张脸
        #绘制人脸网格
        mp_drawing.draw_landmarks(
            image=annotated_image,
            landmark_list=face_landmarks,
            connections=mp_face_mesh.FACEMESH_TESSELATION,
            #landmark_drawing_spec为关键点可视化样式，None为默认样式（不显示关键点）
            # landmark_drawing_spec=mp_drawing_styles.DrawingSpec(thickness=1,circle_radius=2,color=[66,77,229]),
            landmark_drawing_spec=None,
            connection_drawing_spec=mp_drawing_styles.get_default_face_mesh_tesselation_style()
        )
        #绘制人脸轮廓、眼睫毛、眼眶、嘴唇
        mp_drawing.draw_landmarks(
            image=annotated_image,
            landmark_list=face_landmarks,
            connections=mp_face_mesh.FACEMESH_CONTOURS,
            # landmark_drawing_spec为关键点可视化样式，None为默认样式（不显示关键点）
            # landmark_drawing_spec=mp_drawing_styles.DrawingSpec(thickness=1,circle_radius=2,color=[66,77,229]),
            landmark_drawing_spec=None,
            connection_drawing_spec=mp_drawing_styles.get_default_face_mesh_tesselation_style()

        )
        #绘制瞳孔区域
        mp_drawing.draw_landmarks(
            image=annotated_image,
            landmark_list=face_landmarks,
            connections=mp_face_mesh.FACEMESH_IRISES,
            # landmark_drawing_spec为关键点可视化样式，None为默认样式（不显示关键点）
            landmark_drawing_spec=mp_drawing_styles.DrawingSpec(thickness=1,circle_radius=2,color=[128,256,229]),
            # landmark_drawing_spec=None,
            connection_drawing_spec=mp_drawing_styles.get_default_face_mesh_tesselation_style()

        )

cv.imwrite('test.jpg',annotated_image)
look_img(annotated_image)
# 在三维坐标中分别可视化人脸网格、轮廓、瞳孔
mp_drawing.plot_landmarks(results.multi_face_landmarks[0],mp_face_mesh.FACEMESH_TESSELATION)
mp_drawing.plot_landmarks(results.multi_face_landmarks[0],mp_face_mesh.FACEMESH_CONTOURS)
mp_drawing.plot_landmarks(results.multi_face_landmarks[0],mp_face_mesh.FACEMESH_IRISES)

单张图像人脸检测

可以通过调用open3d实现3d模型建立，部分代码与上面类似

import cv2 as cv
import  mediapipe as mp
import numpy as np
from tqdm import tqdm
import time
import  matplotlib.pyplot as plt

# 定义可视化图像函数
def look_img(img):
    img_RGB=cv.cvtColor(img,cv.COLOR_BGR2RGB)
    plt.imshow(img_RGB)
    plt.show()

# 导入三维人脸关键点检测模型
mp_face_mesh=mp.solutions.face_mesh
# help(mp_face_mesh.FaceMesh)

model=mp_face_mesh.FaceMesh(
    static_image_mode=True,#TRUE:静态图片/False:摄像头实时读取
    refine_landmarks=True,#使用Attention Mesh模型
    max_num_faces=40,
    min_detection_confidence=0.2, #置信度阈值，越接近1越准
    min_tracking_confidence=0.5,#追踪阈值
)


# 导入可视化函数和可视化样式
mp_drawing=mp.solutions.drawing_utils
# mp_drawing_styles=mp.solutions.drawing_styles
draw_spec=mp_drawing.DrawingSpec(thickness=2,circle_radius=1,color=[223,155,6])
# 读取图像

img=cv.imread('../人脸三维关键点检测/dkx.jpg')
# width=img1.shape[1]
# height=img1.shape[0]
# img=cv.resize(img1,(width*10,height*10))
# look_img(img)

# 将图像模型输入，获取预测结果

# BGR转RGB
img_RGB=cv.cvtColor(img,cv.COLOR_BGR2RGB)

# 将RGB图像输入模型，获取预测结果

results=model.process(img_RGB)
# # 预测人人脸个数
# len(results.multi_face_landmarks)
#
# print(len(results.multi_face_landmarks))

if results.multi_face_landmarks:
    for face_landmarks  in results.multi_face_landmarks:
        mp_drawing.draw_landmarks(
            image=img,
            landmark_list=face_landmarks,
            connections=mp_face_mesh.FACEMESH_CONTOURS,
            landmark_drawing_spec=draw_spec,
            connection_drawing_spec=draw_spec
        )
else:
    print('未检测出人脸')
look_img(img)
mp_drawing.plot_landmarks(results.multi_face_landmarks[0],mp_face_mesh.FACEMESH_TESSELATION)
mp_drawing.plot_landmarks(results.multi_face_landmarks[1],mp_face_mesh.FACEMESH_CONTOURS)
mp_drawing.plot_landmarks(results.multi_face_landmarks[1],mp_face_mesh.FACEMESH_IRISES)


# 交互式三维可视化
coords=np.array(results.multi_face_landmarks[0].landmark)
# print(len(coords))
# print(coords)

def get_x(each):
    return each.x
def get_y(each):
    return each.y
def get_z(each):
    return each.z

# 分别获取所有关键点的XYZ坐标

points_x=np.array(list(map(get_x,coords)))
points_y=np.array(list(map(get_y,coords)))
points_z=np.array(list(map(get_z,coords)))

# 将三个方向的坐标合并
points=np.vstack((points_x,points_y,points_z)).T
print(points.shape)

import open3d
point_cloud=open3d.geometry.PointCloud()
point_cloud.points=open3d.utility.Vector3dVector(points)
open3d.visualization.draw_geometries([point_cloud])

这是建立的3d的可视化模型，可以通过鼠标拖动将其旋转

摄像头实时关键点检测

定义可视化图像函数
导入三维人脸关键点检测模型
导入可视化函数和可视化样式
读取单帧函数
主要代码和上面的图像类似

import cv2 as cv
import  mediapipe as mp
from tqdm import tqdm
import time
import  matplotlib.pyplot as plt


# 导入三维人脸关键点检测模型
mp_face_mesh=mp.solutions.face_mesh
# help(mp_face_mesh.FaceMesh)

model=mp_face_mesh.FaceMesh(
    static_image_mode=False,#TRUE:静态图片/False:摄像头实时读取
    refine_landmarks=True,#使用Attention Mesh模型
    max_num_faces=5,#最多检测几张人脸
    min_detection_confidence=0.5, #置信度阈值，越接近1越准
    min_tracking_confidence=0.5,#追踪阈值
)


# 导入可视化函数和可视化样式
mp_drawing=mp.solutions.drawing_utils
mp_drawing_styles=mp.solutions.drawing_styles

# 处理单帧的函数

def process_frame(img):
    #记录该帧处理的开始时间
    start_time=time.time()
    img_RGB=cv.cvtColor(img,cv.COLOR_BGR2RGB)
    results=model.process(img_RGB)
    if results.multi_face_landmarks:
        for face_landmarks in results.multi_face_landmarks:
            # mp_drawing.draw_detection(
            #  image=img,
            # landmarks_list=face_landmarks,
            # connections=mp_face_mesh.FACEMESH_TESSELATION,
            # landmarks_drawing_spec=None,
            # landmarks_drawing_spec=mp_drawing_styles.get_default_face_mesh_tesselation_style()
            # )

            # 绘制人脸网格
            mp_drawing.draw_landmarks(
                image=img,
                landmark_list=face_landmarks,
                connections=mp_face_mesh.FACEMESH_TESSELATION,
                # landmark_drawing_spec为关键点可视化样式，None为默认样式（不显示关键点）
                # landmark_drawing_spec=mp_drawing_styles.DrawingSpec(thickness=1,circle_radius=2,color=[66,77,229]),
                landmark_drawing_spec=None,
                connection_drawing_spec=mp_drawing_styles.get_default_face_mesh_tesselation_style()
            )
            # 绘制人脸轮廓、眼睫毛、眼眶、嘴唇
            mp_drawing.draw_landmarks(
                image=img,
                landmark_list=face_landmarks,
                connections=mp_face_mesh.FACEMESH_CONTOURS,
                # landmark_drawing_spec为关键点可视化样式，None为默认样式（不显示关键点）
                # landmark_drawing_spec=mp_drawing_styles.DrawingSpec(thickness=1,circle_radius=2,color=[66,77,229]),
                landmark_drawing_spec=None,
                connection_drawing_spec=mp_drawing_styles.get_default_face_mesh_tesselation_style()

            )
            # 绘制瞳孔区域
            mp_drawing.draw_landmarks(
                image=img,
                landmark_list=face_landmarks,
                connections=mp_face_mesh.FACEMESH_IRISES,
                # landmark_drawing_spec为关键点可视化样式，None为默认样式（不显示关键点）
                # landmark_drawing_spec=mp_drawing_styles.DrawingSpec(thickness=1, circle_radius=2, color=[0, 1, 128]),

                landmark_drawing_spec=None,
                connection_drawing_spec=mp_drawing_styles.get_default_face_mesh_tesselation_style())
    else:
        img = cv.putText(img, 'NO FACE DELECTED', (25 , 50 ), cv.FONT_HERSHEY_SIMPLEX, 1.25,
                         (218, 112, 214), 1, 8)


    #记录该帧处理完毕的时间
    end_time=time.time()
    #计算每秒处理图像的帧数FPS
    FPS=1/(end_time-start_time)
    scaler=1
    img=cv.putText(img,'FPS'+str(int(FPS)),(25*scaler,100*scaler),cv.FONT_HERSHEY_SIMPLEX,1.25*scaler,(0,0,255),1,8)
    return img


# 调用摄像头
cap=cv.VideoCapture(0)

cap.open(0)
# 无限循环，直到break被触发
while cap.isOpened():
    success,frame=cap.read()
    # if not success:
    #     print('ERROR')
    #     break
    frame=process_frame(frame)
    #展示处理后的三通道图像
    cv.imshow('my_window',frame)
    if cv.waitKey(1) &0xff==ord('q'):
        break

cap.release()
cv.destroyAllWindows()