Detect & Mark Face Landmarks Blog

detect-faces.py

# detect face from image using cnn_face_detector
all_face_locations = cnn_face_detector(face_image, 1)

detect-specific-keypoints.py

# imports required libraries
import dlib
import cv2
import numpy as np
from imutils import face_utils

# only for google colab
# from google.colab.patches import cv2_imshow

# initialize dlib face detector and shape predictor
# CNN face detector with mmod_human_face_detector.dat weights
weights = 'models/mmod_human_face_detector.dat'
cnn_face_detector = dlib.cnn_face_detection_model_v1(weights)

predictor_path = 'models/shape_predictor_68_face_landmarks.dat'
shape_predictor = dlib.shape_predictor(predictor_path)

# detect only jawline
def detect_jawline(face_image,keypoints):
  for index,(x, y)  in enumerate(np_keypoints):
    if index in range(0,17): 
      cv2.circle(face_image,(x,y), 3, (0,0,255), -1)
  return face_image

# detect only left eyebrow
def detect_left_eyebrow(face_image,keypoints):
  for index,(x, y)  in enumerate(np_keypoints):
    if index in range(17,22): 
      cv2.circle(face_image,(x,y), 3, (0,0,255), -1)
  return face_image

# detect only right eyebrow
def detect_right_eyebrow(face_image,keypoints):
  for index,(x, y)  in enumerate(np_keypoints):
    if index in range(22,27): 
      cv2.circle(face_image,(x,y), 3, (0,0,255), -1)
  return face_image

# detect only nose
def detect_nose(face_image,keypoints):
  for index,(x, y)  in enumerate(np_keypoints):
    if index in range(27,36): 
      cv2.circle(face_image,(x,y), 3, (0,0,255), -1)
  return face_image

# detect only left eye
def detect_left_eye(face_image,keypoints):
  for index,(x, y)  in enumerate(np_keypoints):
    if index in range(36,42):
      cv2.circle(face_image,(x,y), 3, (0,0,255), -1)
  return face_image

# detect only right eye
def detect_right_eye(face_image,keypoints):
  for index,(x, y)  in enumerate(np_keypoints):
    if index in range(42,48): 
      cv2.circle(face_image,(x,y), 3, (0,0,255), -1)
  return face_image

# detect only lips
def detect_lips(face_image,keypoints):
  for index,(x, y)  in enumerate(np_keypoints):
    if index in range(48,68): 
      cv2.circle(face_image,(x,y), 3, (0,0,255), -1)
  return face_image


if __name__ == "__main__":

  # read image using opencv
  face_image = cv2.imread('images/Barack Obama.png')

  # detect face from image using cnn_face_detector
  all_face_locations = cnn_face_detector(face_image, 1)

  # loop over every detected face
  for index, face_location in enumerate(all_face_locations):

    # pass the face image along with location to shape predictor
    keypoints = shape_predictor(face_image, face_location.rect)
    np_keypoints = face_utils.shape_to_np(keypoints)

    # detect only left eye
    face_image = detect_left_eye(face_image, np_keypoints)

  # dispaly results
  cv2.imshow("Face Landmarks", face_image)

  # only for google colab
  # cv2_imshow(face_image)

dir-tree.md

•
├── images
│   ├── Barack Obama.png
│   └── Barack Obama 2.png
├── models
│   ├── shape_predictor_68_face_landmarks.dat
│   └── mmod_human_face_detector.dat
├── facial-landmarks-using-dlib.py
├── facial-landmarks-using-face-recognition.py
└── detect-specific-keypoints.py

display-result.py

# show the results
cv2.imshow("Face Landmarks",face_image)
    
# only for google colab
# cv2_imshow(face_image)

dlib-landmarks.py

# importing required libraries
import  dlib
import cv2
from imutils import face_utils

# only for google colab
# from google.colab.patches import cv2_imshow

# read image using opencv
face_image = cv2.imread('images/Barack Obama.png')

# initialize dlib face detector and shape predictor
# CNN face detector with mmod_human_face_detector.dat weights
weights = 'models/mmod_human_face_detector.dat'
cnn_face_detector = dlib.cnn_face_detection_model_v1(weights)

predictor_path = 'models/shape_predictor_68_face_landmarks.dat'
shape_predictor = dlib.shape_predictor(predictor_path)

# detect face from image using cnn_face_detector
all_face_locations = cnn_face_detector(face_image, 1)

# loop over every detected face
for index, face_location in enumerate(all_face_locations):

  # pass the face image along with location to shape predictor
  keypoints = shape_predictor(face_image, face_location.rect)
  # convert keypoints to numpy array
  np_keypoints = face_utils.shape_to_np(keypoints)
  
  # marks each keypoint detected on the face
  for index,(x, y)  in enumerate(np_keypoints):
    cv2.circle(face_image,(x,y), 3, (0,0,255), -1)
    font = cv2.FONT_HERSHEY_SIMPLEX
    # number each point
    cv2.putText(face_image, str(index), (x+2,y), font, 0.5, (0,255,0), 1)

# show the results
cv2.imshow("Face Landmarks",face_image)
    
# only for google colab
# cv2_imshow(face_image)

import-libraries.py

# importing required libraries
import  dlib
import cv2
from imutils import face_utils

# only for google colab
# from google.colab.patches import cv2_imshow

initialize-model.py

# initialize dlib face detector and shape predictor
# CNN face detector with mmod_human_face_detector.dat weights
weights = 'models/mmod_human_face_detector.dat'
cnn_face_detector = dlib.cnn_face_detection_model_v1(weights)

predictor_path = 'models/shape_predictor_68_face_landmarks.dat'
shape_predictor = dlib.shape_predictor(predictor_path)

logic.py

# loop over every detected face
for index, face_location in enumerate(all_face_locations):

  # pass the face image along with location to shape predictor
  keypoints = shape_predictor(face_image, face_location.rect)
  # convert keypoints to numpy array
  np_keypoints = face_utils.shape_to_np(keypoints)
  
  # marks each keypoint detected on the face
  for index,(x, y)  in enumerate(np_keypoints):
    cv2.circle(face_image,(x,y), 3, (0,0,255), -1)
    font = cv2.FONT_HERSHEY_SIMPLEX
    # number each point
    cv2.putText(face_image, str(index), (x+2,y), font, 0.5, (0,255,0), 1)

mark-face-landmarks.py

# importing required libraries
import  face_recognition
from PIL import Image, ImageDraw

# Load the jpg file into a numpy array
face_image = face_recognition.load_image_file("images/Barack Obama.png")

# get the face landmarks list
face_landmarks_list = face_recognition.face_landmarks(face_image)

# print all the face landmarks
print(face_landmarks_list)

# for loop tointerate through all land marks in the list
for face_landmarks in face_landmarks_list:

  # convert numpy array to pil image and create a Draw Object  
  pil_image = Image.fromarray(face_image)
  d = ImageDraw.Draw(pil_image)

  # draw green line connecting the dots
  d.line(face_landmarks['chin'], fill=(0,255,0), width=2)
  d.line(face_landmarks['left_eyebrow'], fill= (0,255,0), width=2)
  d.line(face_landmarks['right_eyebrow'], fill= (0,255,0), width=2)
  d.line(face_landmarks['nose_bridge'], fill= (0,255,0), width=2)
  d.line(face_landmarks['nose_tip'], fill= (0,255,0), width=2)
  d.line(face_landmarks['right_eye'], fill= (0,255,0), width=2)
  d.line(face_landmarks['left_eye'], fill= (0,255,0), width=2)
  d.line(face_landmarks['top_lip'], fill= (0,255,0), width=2)
  d.line(face_landmarks['bottom_lip'], fill= (0,255,0), width=2)

# display the image
pil_image.show()

# You can also asave a copy of the new image to diskk
pil_image.save("Face_With_Landmarks.jpg")

read-image.py

# read image using opencv
face_image = cv2.imread('images/Barack Obama.png')