Skip to content

A demonstration of face database search implemented in python

License

Notifications You must be signed in to change notification settings

lehgtrung/face-search

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

26 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

face-search

This is a demonstration of a face search system

Work flows:

  • Detect faces from the images
  • Crop the face to 224x224 pixels
  • Align the face
  • Feed face images through a feature extractor (I use vgg face feature extractor)
  • Save the vector representations into database along with images names and url
  • Queries run very fast thanks to Postgres's CUBE extension

This demo works on Ubuntu 16.04 LTS with Python 2.7, my laptop details:

  • CPU: Intel core i5-4210U @1.7GHz
  • RAM: 8GB DDR3
  • GPU: NVIDIA GTX 840M

Installation

OpenCV for python

Details instruction: https://gist.github.com/trungkak/a934e92b3829a025f98a0b3419fad2da

CUDA, tensorflow, keras installation

Details instruction: https://gist.github.com/trungkak/3d9d0c1b9dda91623488a5e4a3373053

Postgres installation

Details instruction: https://www.digitalocean.com/community/tutorials/how-to-install-and-use-postgresql-on-ubuntu-16-04

Preparation

Get into src directory

cd /src

Open .env file then enter your config (DBName, USER, PASSWORD)

vi .env

To do things from scratch, download a face dataset like LFW (http://vis-www.cs.umass.edu/lfw/)

cd ../images
wget http://vis-www.cs.umass.edu/lfw/lfw-funneled.tgz
tar zxvf lfw-funneled.tgz

Running

To index your images data, use:

python app.py -path <path-to-your-images-directory>

To identify a person, use:

python app.py -image <path-to-your-image>

To get help, use:

python app.py -h

Resources

  • Deep Face Recognition: Link
  • Face search at scale, 80 million gallery: Link

Author

Releases

No releases published

Packages

No packages published

Languages