人脸识别调研

项目列表

开源项目	语言/环境	准确率 (on Labeled Faces in the Wild)	备注
ageitgey/face_recognition	Python, Dlib	99.38%	Docs
DFace	Python
facenet	TensorFlow	99.2%	https://mp.weixin.qq.com/s/1kgbYScIujSjCRvfPGw0tg
VGG Face Descriptor	Caffe, Torch	99.13%
SeetaFace Engine	C++	97.1%	中科院山世光
openbiometrics	C/C++

其他

https://skybiometry.com/pricing/

https://www.betafaceapi.com/wpa/index.php/pricing

https://facedetection.com/software/

人脸识别大赛

http://www.msceleb.org/leaderboard/iccvworkshop-c1

ageitgey/face_recognition 项目调研

项目地址

https://github.com/ageitgey/face_recognition

 

性能

99.38% on the Labeled Faces in the Wild benchmark

 

环境

macOS/Linux

Python 3.3+/Python 2.7

dlib(a C++ ML toolkit, https://github.com/davisking/dlib)

 

Python Module API Docs

https://face-recognition.readthedocs.io

 

安装

git clone 
https://github.com/davisking/dlib.git

sudo apt-get install cmake
sudo apt-get install libboost-python-dev
sudo python setup.py install
pip2 install face_recognition

 

使用 (Recognize faces)

 

import face_recognition
 
picture_of_me = face_recognition.load_image_file("me.jpg")
my_face_encoding = face_recognition.face_encodings(picture_of_me)[0]
 
# my_face_encoding now contains a universal 'encoding' of my facial features that can be compared to any other picture of a face!
 
unknown_picture = face_recognition.load_image_file("unknown.jpg")
unknown_face_encoding = face_recognition.face_encodings(unknown_picture)[0]
 
results = face_recognition.compare_faces([my_face_encoding], unknown_face_encoding)

 

总结：

1. 后续需要解决如何部署在服务器上，Docker？

2. 1:1 识别性能很好

3. 1:N 识别需要事先注册并存储face_encodings，然后遍历比对，性能有待测试