Facial Landmark Detection

1. Course Content

• Run the example program to detect the face using the mediapipe framework.

2. Program Startup

• Run the face detection node

ros2 launch yahboomcar_bringup mediapipe_demo.launch.py demo:=FaceMesh

After the program runs, as shown in the figure below, the detected facial points will be displayed on the right side of the image.

3. Source Code Analysis

• Program code path:

xxxxxxxxxx
~/yahboomcar_ros2_ws/yahboomcar_ws/src/common_demo/common_demo/mediapipe/04_FaceMesh.py

Import the required library files,

xxxxxxxxxx
import rclpy
from rclpy.node import Node
#Import the mediapipe library
import mediapipe as mp
import cv2 as cv
import numpy as np
import time
import os
from cv_bridge import CvBridge
from sensor_msgs.msg import Image
from arm_msgs.msg import ArmJoints
import cv2
print("import done")

Initialize the data and define the Publisher and Subscriber,

xxxxxxxxxx
def __init__(self, name,staticMode=False, maxFaces=2, minDetectionCon=0.5, minTrackingCon=0.5):
    super().__init__(name)
    self.mpDraw = mp.solutions.drawing_utils
    #Use the class in the mediapipe library to define a face object
    self.mpFaceMesh = mp.solutions.face_mesh
    self.faceMesh = self.mpFaceMesh.FaceMesh(
    static_image_mode=staticMode,
    max_num_faces=maxFaces,
    min_detection_confidence=minDetectionCon,
    min_tracking_confidence=minTrackingCon )
    #Define the properties of the joint connection lines, which will be used in the subsequent joint point connection function
    self.lmDrawSpec = mp.solutions.drawing_utils.DrawingSpec(color=(0, 0, 255), thickness=-1, circle_radius=3)
    self.drawSpec = self.mpDraw.DrawingSpec(color=(0, 255, 0), thickness=1, circle_radius=1)
    self.rgb_bridge = CvBridge()
    #Define the Subscriber for the color image topic
    self.sub_rgb = self.create_subscription(Image,"/camera/color/image_raw",self.get_RGBImageCallBack,100)

Color image callback function,

xxxxxxxxxx
def get_RGBImageCallBack(self,msg):
    #Use CvBridge to convert the color image message data into image data
    rgb_image = self.rgb_bridge.imgmsg_to_cv2(msg, "bgr8")
    #Pass the obtained image into the defined pubFaceMeshPoint function; draw=False means not drawing the joint points on the original color image
    frame, img = self.pubFaceMeshPoint(rgb_image, draw=False)
    #Merge the two images
    dist = self.frame_combine(frame, img)
    key = cv2.waitKey(1)
    cv.imshow('dist', dist)

pubFaceMeshPoint function,

xxxxxxxxxx
def pubFaceMeshPoint(self, frame, draw=True):
    #Create a new image based on the size of the input image; the data type of the image is uint8
    img = np.zeros(frame.shape, np.uint8)
    #Convert the color space of the input image from BGR to RGB for subsequent image processing
    imgRGB = cv.cvtColor(frame, cv.COLOR_BGR2RGB)
    #Call the process function in the mediapipe library for image processing; during init, the self.faceMesh object was created and initialized
    self.results = self.faceMesh.process(imgRGB)
    #Determine whether self.results.multi_face_landmarks exists, that is, whether a face has been recognized
    if self.results.multi_face_landmarks:
        for i in range(len(self.results.multi_face_landmarks)):
            if draw: self.mpDraw.draw_landmarks(frame, self.results.multi_face_landmarks[i], self.mpFaceMesh.FACEMESH_CONTOURS, self.lmDrawSpec, self.drawSpec)
            #Connect each joint point on the blank image created earlier
            self.mpDraw.draw_landmarks(img, self.results.multi_face_landmarks[i], self.mpFaceMesh.FACEMESH_CONTOURS, self.lmDrawSpec, self.drawSpec)
    return frame, img

The frame_combine image merging function was mentioned in the first lesson of this chapter; refer to the analysis of this function in "1. Hand Detection".