最近在训练一个人脸识别的模型,而项目训练需要大量真实人脸图片样本。
刚好项目用到opencv识别人脸,可以把每一帧图片保存下来,用此方法可以方便的获取大量的脸部样本,大约20分钟可以获取到10000张.
#-*- encoding:utf8 -*- import cv2 import os import sys import random # 获取分类器 classifier = cv2.CascadeClassifier('haarcascade_frontalface_default.xml') face_dir = './my_faces' if not os.path.exists(face_dir): os.makedirs(face_dir) name=raw_input("please input your name:") os.makedirs(face_dir+'/'+name) # 打开摄像头 参数为输入流,可以为摄像头或视频文件 camera = cv2.VideoCapture(0) # 改变亮度与对比度 def relight(img, alpha=1, bias=0): w = img.shape[1] h = img.shape[0] #image = [] for i in range(0,w): for j in range(0,h): for c in range(3): tmp = int(img[j,i,c]*alpha + bias) if tmp > 255: tmp = 255 elif tmp < 0: tmp = 0 img[j,i,c] = tmp return img i = 1 while 1: if (i <= 10000): print('It`s processing %s image.' % i) success, img = camera.read() # 读帧 gray_img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)#转灰度图 faces = classifier.detectMultiScale(gray_img, 1.3, 5)#用分类器获取脸部 for f_x, f_y, f_w, f_h in faces:#截取原来图像的脸部 face = img[f_y:f_y+f_h, f_x:f_x+f_w] face = cv2.resize(face, (128,128)) face = relight(face, random.uniform(0.5, 1.5), random.randint(-50, 50))#重新调亮度 cv2.imwrite(face_dir+'/'+name+'/'+str(i)+'.jpg', face) i+=1 key = cv2.waitKey(10) c = chr(key & 255) if c in ['q', 'Q', chr(27)]: break else: break