Face tracking

1. Experimental purpose

Drive the car to do face tracking

2. Experimental path source code

Enter the car system, end the car program, enter "ip (ip is the car's ip): 8888" in the browser, enter the password "yahboom" Then log in Enter the path /home/pi/Rider-pi_class/6.AI Visual Interaction Course/4. Facial tracking/ and run face_decetion.ipynb. Or directly enter

/home/pi/Rider-pi_class/6.AI Visual Interaction Course/4. Facial tracking/
python3 face_decetion.py

3. Experimental phenomenon

After running the source code, the car tracks the face detected by the camera. The car cannot be in a stopped state (i.e. not in a standing balance state), otherwise it cannot move.

4. Analysis of main source code parameters

​x
#-----------------------COMMON INIT-----------------------
import cv2
import mediapipe as mp
mp_face_detection = mp.solutions.face_detection
mp_drawing = mp.solutions.drawing_utils
​
# For static images:
IMAGE_FILES = []
with mp_face_detection.FaceDetection(
    model_selection=1, min_detection_confidence=0.5) as face_detection:
  for idx, file in enumerate(IMAGE_FILES):
    image = cv2.imread(file)
    # Convert the BGR image to RGB and process it with MediaPipe Face Detection.
    results = face_detection.process(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
​
    # Draw face detections of each face.
    if not results.detections:
      continue
    annotated_image = image.copy()
    for detection in results.detections:
      print('Nose tip:')
      print(mp_face_detection.get_key_point(
          detection, mp_face_detection.FaceKeyPoint.NOSE_TIP))
      mp_drawing.draw_detection(annotated_image, detection)
    cv2.imwrite('/tmp/annotated_image' + str(idx) + '.png', annotated_image)
​
# For webcam input:
cap=cv2.VideoCapture(0)
cap.set(3,320)
cap.set(4,240)
with mp_face_detection.FaceDetection(
    model_selection=0, min_detection_confidence=0.5) as face_detection:
  while cap.isOpened():
    success, image = cap.read()
    if not success:
      print("Ignoring empty camera frame.")
      # If loading a video, use 'break' instead of 'continue'.
      continue
​
    # To improve performance, optionally mark the image as not writeable to
    # pass by reference.
    image.flags.writeable = False
    image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
    results = face_detection.process(image)
​
    # Draw the face detection annotations on the image.
    image.flags.writeable = True
    image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
    if results.detections:
      for detection in results.detections:
        value_x=0
        value_y=0
        mp_drawing.draw_detection(image, detection)
        xy=(mp_face_detection.get_key_point(detection, mp_face_detection.FaceKeyPoint.NOSE_TIP))
        face_x=320-xy.x*320
        face_y=xy.y*240
        value_x = face_x - 160
        value_y = face_y - 120
        rider_x=value_x
        print(face_x,face_y)
        if value_x > 55:
          value_x = 55
        elif value_x < -55:
          value_x = -55
        if value_y > 75:
          value_y = 75
        elif value_y < -75:
          value_y = -75
          
    else:
      value_x=value_y=face_x=face_y=0
      rider_x=9999
    print(['y','p'],[value_x/9, value_y/15])
    if car_type=='L' or car_type=='M':
      car.attitude(['y','p'],[value_x/9, value_y/15])
    elif car_type=='R':
      print(value_x,value_y)
      car.rider_height(75+int((190-face_y)/70*40))
      if rider_x==9999:
        car.rider_turn(0)
      else:
        if rider_x > 35:
          car.rider_turn(20)
        elif rider_x < -35:
          car.rider_turn(-20)
        else:
          car.rider_turn(0)
    b,g,r = cv2.split(image)
    image = cv2.merge((r,g,b))
    image = cv2.flip(image, 1)
    imgok = Image.fromarray(image)
    display.ShowImage(imgok)
​
    #显示在屏幕上
    r,g,b = cv2.split(image)
    image1 = cv2.merge((b,g,r))
    cv2.imshow("image1",image1)
​
    # Flip the image horizontally for a selfie-view display.
    #cv2.imshow('MediaPipe Face Detection', cv2.flip(image, 1))
    if cv2.waitKey(5) & 0xFF == 27:
      break
    if button.press_b():
      car.reset()
      break
cap.release()
​

From the source code analysis, we can get that the car will track the face according to the camera. If the face looks up, the car will stand up; if the face looks down, the car will squat; if the face looks left or right, the car will follow.