注释语句

README.md

@@ -0,0 +1 @@
+利用OpenCV、CNN进行人脸识别

get_img_from_camera.py

@@ -0,0 +1,72 @@
+#!/bin/python
+# -*- coding: utf8 -*-
+# __author__ = '尹籽'
+
+import cv2
+import dlib
+import os
+import sys
+import random
+
+output_dir = './faces_creat/222'
+size = 200
+
+if not os.path.exists(output_dir):
+ os.makedirs(output_dir)
+
+# 改变图片的亮度与对比度
+def relight(img, light=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]*light + bias)
+ if tmp > 255:
+ tmp = 255
+ elif tmp < 0:
+ tmp = 0
+ img[j,i,c] = tmp
+ return img
+
+#使用dlib自带的frontal_face_detector作为我们的特征提取器
+detector = dlib.get_frontal_face_detector()
+# 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)
+
+index = 1
+while True:
+ if (index <= 600):
+ print('Being processed picture %s' % index)
+ # 从摄像头读取照片
+ success, img = camera.read()
+ # 转为灰度图片
+ gray_img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+ # 使用detector进行人脸检测
+ dets = detector(gray_img, 1)
+
+ for i, d in enumerate(dets):
+ x1 = d.top() if d.top() > 0 else 0
+ y1 = d.bottom() if d.bottom() > 0 else 0
+ x2 = d.left() if d.left() > 0 else 0
+ y2 = d.right() if d.right() > 0 else 0
+
+ face = img[x1:y1,x2:y2]
+ # 调整图片的对比度与亮度， 对比度与亮度值都取随机数，这样能增加样本的多样性
+ face = relight(face, random.uniform(0.5, 1.5), random.randint(-50, 50))
+
+ face = cv2.resize(face, (size,size))
+ gray= cv2.cvtColor(face, cv2.COLOR_BGR2GRAY)
+
+ cv2.imshow('image', face)
+
+ cv2.imwrite(output_dir+'/'+str(index)+'l.jpg', gray)
+
+ index += 1
+ key = cv2.waitKey(30) & 0xff
+ if key == 27:
+ break
+ else:
+ print('Finished!')
+ break

get_img_from_video.py

@@ -0,0 +1,68 @@
+import cv2
+import dlib
+import os
+import sys
+import random
+
+output_dir = './faces_creat/xueru'
+size = 200
+
+if not os.path.exists(output_dir):
+ os.makedirs(output_dir)
+
+# 改变图片的亮度与对比度
+def relight(img, light=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]*light + bias)
+ if tmp > 255:
+ tmp = 255
+ elif tmp < 0:
+ tmp = 0
+ img[j,i,c] = tmp
+ return img
+
+#使用dlib自带的frontal_face_detector作为我们的特征提取器
+detector = dlib.get_frontal_face_detector()
+# 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)
+
+index = 1
+while True:
+ if (index <= 600):
+ print('Being processed picture %s' % index)
+ # 从摄像头读取照片
+ success, img = camera.read()
+ # 转为灰度图片
+ gray_img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+ # 使用detector进行人脸检测
+ dets = detector(gray_img, 1)
+
+ for i, d in enumerate(dets):
+ x1 = d.top() if d.top() > 0 else 0
+ y1 = d.bottom() if d.bottom() > 0 else 0
+ x2 = d.left() if d.left() > 0 else 0
+ y2 = d.right() if d.right() > 0 else 0
+
+ face = img[x1:y1,x2:y2]
+ # 调整图片的对比度与亮度， 对比度与亮度值都取随机数，这样能增加样本的多样性
+ face = relight(face, random.uniform(0.5, 1.5), random.randint(-50, 50))
+
+ face = cv2.resize(face, (size,size))
+ gray= cv2.cvtColor(face, cv2.COLOR_BGR2GRAY)
+
+ cv2.imshow('image', face)
+
+ cv2.imwrite(output_dir+'/'+str(index)+'.jpg', gray)
+
+ index += 1
+ key = cv2.waitKey(30) & 0xff
+ if key == 27:
+ break
+ else:
+ print('Finished!')
+ break