使用MediaPipe Python进行人脸识别

MediaPipe介绍

MediaPipe 是一款由 Google Research 开发并开源的多媒体机器学习模型应用框架。在谷歌，一系列重要产品，如 、Google Lens、ARCore、Google Home 以及 ，都已深度整合了 MediaPipe。

MediaPipe安装

我使用Ubuntu 20.04，Python 3.8 可以直接在命令行使用pip安装。

pip install opencv-contrib-python
pip install mediapipe

MediaPipe人脸识别Python代码

我使用的测试视频是618c1d1b9d4a407096f5a287_lookV5-transcode.mp4，也可以尝试webm格式的视频。

import cv2
import mediapipe as mp
import time


cap = cv2.VideoCapture('./618c1d1b9d4a407096f5a287_lookV5-transcode.mp4')
pTime = 0

mpDraw = mp.solutions.drawing_utils
mpFaceMesh = mp.solutions.face_mesh
faceMesh = mpFaceMesh.FaceMesh(max_num_faces=2)
drawSpec = mpDraw.DrawingSpec(thickness=1, circle_radius=0)

w = cap.get(cv2.CAP_PROP_FRAME_WIDTH)
h = cap.get(cv2.CAP_PROP_FRAME_HEIGHT)
fps = cap.get(cv2.CAP_PROP_FPS)
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
out = cv2.VideoWriter('./618c1d1b9d4a407096f5a287_lookV5-transcode-save.mp4', fourcc, fps, (int(w), int(h)), True)

c = 0
i = 0
timeF = 3

while (cap.isOpened()):
    success, img = cap.read()
    imgRGB = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
    results = faceMesh.process(imgRGB)
    if results.multi_face_landmarks:
        for faceLms in results.multi_face_landmarks:
            mpDraw.draw_landmarks(img, faceLms, mpFaceMesh.FACEMESH_TESSELATION, drawSpec, drawSpec)

    cTime = time.time()
    fps = 1 / (cTime - pTime)
    pTime = cTime
    cv2.putText(img, f'ZRT face puppet, in realtime Unreal Engine 4 (4k @ 60fps)', (20, 40), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (52, 124, 90), 1)
    cv2.putText(img, f'FPS: {int(fps)}', (20, 60), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (52, 124, 90), 1)
    # out.write(img)
    c = c + 1
    if (c % timeF == 0) and success:
        out = cv2.resize(img, (0, 0), fx=0.5, fy=0.5)
        cv2.imwrite(str(i) + '.png', out)
        i = i + 1
    cv2.imshow('Face Tracking', img)
    if (cv2.waitKey(5) & 0xFF) == ord('q'):
        break