opencv处理换脸操作

Logout 用cv2处理换脸操作 Last Checkpoint: 06/12/2018 (autosaved)

Python 3 

Not Trusted

• File
• Edit
• View
• Insert
• Cell
• Kernel
• Widgets
• Help

CodeMarkdownRaw NBConvertHeading-

In [18]:

import cv2

# opencv是计算机视觉库

import matplotlib.pyplot as plt

%matplotlib inline 

# 这个魔法指令在程序的结束加上一个plt.show()

In [28]:

an = plt.imread("./cv2_change_head/安倍.jpg")

an

Out[28]:

array([[[187, 186, 200],
        [187, 186, 200],
        [187, 186, 200],
        ..., 
        [185, 183, 196],
        [186, 184, 197],
        [186, 184, 197]],

       [[187, 186, 200],
        [187, 186, 200],
        [187, 186, 200],
        ..., 
        [185, 183, 196],
        [186, 184, 197],
        [186, 184, 197]],

       [[187, 186, 200],
        [187, 186, 200],
        [187, 186, 200],
        ..., 
        [185, 183, 196],
        [186, 184, 197],
        [186, 184, 197]],

       ..., 
       [[ 19,  18,  24],
        [ 18,  17,  23],
        [ 17,  16,  22],
        ..., 
        [ 25,  24,  30],
        [ 24,  23,  29],
        [ 22,  21,  27]],

       [[ 18,  17,  23],
        [ 18,  17,  23],
        [ 16,  15,  21],
        ..., 
        [ 25,  24,  30],
        [ 25,  24,  30],
        [ 21,  20,  26]],

       [[ 17,  16,  22],
        [ 16,  15,  21],
        [ 15,  14,  20],
        ..., 
        [ 25,  24,  30],
        [ 25,  24,  30],
        [ 20,  19,  25]]], dtype=uint8)

In [29]:

plt.imshow(an)

Out[29]:

<matplotlib.image.AxesImage at 0x1bd3e788da0>

In [30]:

# 创建一个cv2的分类器，用于识别图片

case = cv2.CascadeClassifier()

In [31]:

# 给case加上人脸识别的算法

case.load("./cv2_change_head/haarcascade_frontalface_default.xml")

# 加上这个算法以后，case就可以识别人脸的位置

Out[31]:

True

In [32]:

# 用case来识别人脸

face = case.detectMultiScale(an)

face

Out[32]:

array([[325, 278, 504, 504]], dtype=int32)

In [33]:

dog = plt.imread("./cv2_change_head/dog.jpg")

plt.imshow(dog)

Out[33]:

<matplotlib.image.AxesImage at 0x1bd3f4dd278>

In [34]:

case.detectMultiScale(dog)

Out[34]:

()

In [35]:

dog.shape

Out[35]:

(450, 450, 3)

In [39]:

big_dog = cv2.resize(dog,(504,504))

plt.imshow(big_dog)

Out[39]:

<matplotlib.image.AxesImage at 0x1bd43c51668>

In [40]:

an[325:829,278:782] = big_dog

​

In [42]:

plt.imshow(an)

Out[42]:

<matplotlib.image.AxesImage at 0x1bd3f5766d8>