junleen / face-alignment

A easy tool of face alignment based on dlib but with very good interface

68 Facial Landmarks Shape Predictor(Easy Use)

This repo aims at shortening your developing time in face detection and alignment.

I use dlib to perform face detection and landmarks regression, and you can refer to https://github.com/codeniko/shape_predictor_81_face_landmarks for 81 face landmarks detection.

Prepare Dlib model(official release)

Please download the pre-trained model from shape_predictor_68_face_landmarks.dat and extract it to dlibmodels directory.

Requirements

opencv-python>=3.2.0
dlib>=19.14.0
numpy

Usage

In this work, a easy-use face detector is build. refer to dlibpp.py. To detect from a image, just run:

fd = dlibpp.DLIBPP()
box, landmarks = fd.detect_faces(img)

Example code

from dlibpp import DLIBPP
import cv2
import mask

fd = DLBPP()
img = cv2.imread('example.png')
box, landmarks = fd.detect_faces(img)
linemask = np.zeros_like(img).astype(np.uint8)

for landmarks_i in landmarks:
    for (x,y) in landmarks_i:
        img = cv2.circle(img, (x, y), 1, (0,255,0), 1)
	linemask = mask.generate_line_mask(linemask, landmarks_i)
example = np.hstack([img, linemask])
cv2.imshow('example', example)
cv2.waitKey(0)
cv2.destroyAllWindows()

Examples

Acknowledgement

• Thanks for open source code from Dlib
• Thanks for codeniko for his fantastic 81 landmarks alignment work.

Thanks for your interest in this repo. If you work or research benefit from this repo, please cite it , start it and popularize it 😃