顔のランドマーク検出とドロネー分割 Python + OpenCV + dlib

概要

Python + OpenCV + dlib を用いて顔にドロネー図を描画するサンプルを紹介します。

おまけで、カメラから取得した顔画像に描画するサンプルを追加しました。

環境

Python 3.7.0 OpenCV 3.4

デモ

サンプルソース

import cv2
import dlib
import numpy as np

PREDICTOR_PATH = "/path/to/shape_predictor_68_face_landmarks.dat"
predictor = dlib.shape_predictor(PREDICTOR_PATH)

cascade_path='/path/to/haarcascade_frontalface_default.xml'
cascade = cv2.CascadeClassifier(cascade_path)

def get_landmarks(im):
    rects = cascade.detectMultiScale(im, 1.3,5)
    (x,y,w,h) = rects[0]
    rect = dlib.rectangle(x,y,x+w,y+h)
    return np.matrix([[p.x, p.y] for p in predictor(im, rect).parts()])

def annotate_landmarks(im, landmarks):
    im = im.copy()
    for idx, point in enumerate(landmarks):
        pos = (point[0, 0], point[0, 1])
        cv2.putText(im, str(idx), pos,
            fontFace=cv2.FONT_HERSHEY_SCRIPT_SIMPLEX,
            fontScale=0.4,
            color=(255, 0, 0))
        cv2.circle(im, pos, 2, color=(255, 255, 0))
    return im

def rect_contains(rect, point) :
    if point[0] < rect[0] :
        return False
    elif point[1] < rect[1] :
        return False
    elif point[0] > rect[2] :
        return False
    elif point[1] > rect[3] :
        return False
    return True

def draw_delaunay(img, subdiv, delaunay_color ) :

    triangleList = subdiv.getTriangleList()
    size = img.shape
    r = (0, 0, size[1], size[0])

    for t in triangleList :

        pt1 = (t[0], t[1])
        pt2 = (t[2], t[3])
        pt3 = (t[4], t[5])

        if rect_contains(r, pt1) and rect_contains(r, pt2) and rect_contains(r, pt3) :

            cv2.line(img, pt1, pt2, delaunay_color, 1, 16, 0)
            cv2.line(img, pt2, pt3, delaunay_color, 1, 16, 0)
            cv2.line(img, pt3, pt1, delaunay_color, 1, 16, 0)

if __name__ == "__main__" :
    im = cv2.imread('lenna.png')

    height = im.shape[0]
    width = im.shape[1]

    im = cv2.resize(im, (width, height))

    rect = (0, 0, width, height)

    subdiv = cv2.Subdiv2D(rect)

    points = get_landmarks(im)

    for p in points :
        a = np.array(p).ravel()
        subdiv.insert((a[0], a[1]))

    draw_delaunay(im, subdiv, (255, 255, 255) )

    # im = annotate_landmarks(im, points)

    cv2.imshow('result', im)

    cv2.waitKey(0)
    cv2.destroyAllWindows()

実行結果

リアルタイムでカメラ画像を利用する場合

この場合は、メイン部分をいかに書き換えます。

if __name__ == "__main__" :
    cap = cv2.VideoCapture(0)

    while(True):
        try:
            ret, im = cap.read()

            height = im.shape[0] 
            width = im.shape[1] 
            im = cv2.resize(im, (width, height))

            rect = (0, 0, width, height)

            subdiv = cv2.Subdiv2D(rect)

            points = get_landmarks(im)

            for p in points :
                a = np.array(p).ravel()
                subdiv.insert((a[0], a[1]))

            draw_delaunay( im, subdiv, (255, 255, 255) )

            # im = annotate_landmarks(im, points)

            cv2.imshow('Result',im)
            
        except:
            print('error!')        

        if cv2.waitKey(0) & 0xFF == ord('q'):
            break

    cv2.destroyAllWindows()