[PYTHON] I tried using face

I found a face recognition article and tried it, but it didn't go smoothly, so I'll write it down.

Face recognition can be performed using Python. Moreover, the library is open to the public. Moreover, it seems easy. First article I saw: https://www.cresco.co.jp/blog/entry/9468/

I'm new to Python in the first place, but I'm looking at it sweetly because it works even if I write it properly.
Environment is Windows 10, Python 3.8

I was wondering if face recognition could be used for things that are just a problem in my work. I first saw it with a light feeling like "just install it with pip".

＞pip3 install face_recognition

I got an error during installation and couldn't proceed at all ... I googled while looking at the unreadable English

I was angry because I didn't have a wheel. what is that? It seems to be a zip for python distribution.

＞pip install -U wheel

Still more errors dlib

In the first place, face_recognition seems to be for using a face recognition library called dlib. Is it a wrapper for python? Reference: https://qiita.com/nonbiri15/items/f95b5fb01ae38980c9ce

It seems that you need to install dlib.

＞pip3 install dlib

After all error rolls out

It seems that dlib is made in C It seems that CMAKE is necessary. cmake-3.18.0-rc2-win64-x64.msi ↑ DL and install

Still more ~ It seems to build with VC ++ (that's right) Reference: https://qiita.com/strv13570/items/a0542600532deee61391 I had Visual studio 2017, but I didn't want to touch VC ++, so I didn't. Since there is no help for it, change it to use VC ++ and install it

As mentioned above, set the path etc. as appropriate (confirmation)

＞pip3 install dlib
 ・
 ・
 ・
Successfully installed dlib-19.20.0

Finally, I passed dlib.

pip3 install face_recognition
 ・
 ・
 ・
 Successfully installed Pillow-7.1.2 face-recognition-1.3.0 numpy-1.19.0

It was a long way to finally install it ... I thought Gugu, why did everyone put in anacondas? ?? Is it business?

For the time being, create two folders with reference to the site and put photos in it. How to use it like creating a face folder you know and a face folder you don't know, putting them in each, and taking a diff. I tried each below.

I put one real piece in it. Execution result tosi3 has been deceived.

As mentioned above, the intention of the usage review has become an installation memo ...

I plan to try various other things Reference: https://www.kkaneko.jp/dblab/dlib/facerec.html

Postscript: ↑ I looked at the site and tried it

import face_recognition

src_img = face_recognition.load_image_file("./face/obama.png ")
src_img_encoding = face_recognition.face_encodings(src_img)[0]
 print(src_img_encoding)

img1="./face2/tosi.png "
img2="./face2/tosi2.png "
img3="./face2/tosi3.png "

dest_img1 = face_recognition.load_image_file(img1)
dest_img_encoding1 = face_recognition.face_encodings(dest_img1)[0]

dest_img_encoding2 = face_recognition.face_encodings(face_recognition.load_image_file(img2))[0]
dest_img_encoding3 = face_recognition.face_encodings(face_recognition.load_image_file(img3))[0]

results = face_recognition.compare_faces([src_img_encoding], dest_img_encoding1)
print(img1 , results)
results = face_recognition.compare_faces([src_img_encoding], dest_img_encoding2)
print(img2 , results)
results = face_recognition.compare_faces([src_img_encoding], dest_img_encoding3)
print(img3 , results)

＞test_face.py
./face2/tosi.png [False]
./face2/tosi2.png [False]
./face2/tosi3.png [True]

Process finished with exit code 0

Facial features can be quantified and compared It ’s really easy! !!

[PYTHON] I tried using face_recognition