kelvinjjwong / face-club Goto Github PK

A web application help recognize and maintenance faces from images imported by ImageDocker

1. scan picture files from a mounted volume
2. detect and recognize faces from the pictures
3. store the names of recognized faces into database
4. provide front-end to tag and untag faces on pictures

• face detection: dlib, opencv, face_recognition
• web: flask
• db: sqlalchemy(sqlite3), asyncpg
• schedule: apscheduler

• Will connect to a PostgreSQL database maintained by ImageDocker
• Will maintain a SQLite database locally

1. Install Anaconda for mac
2. Install Homebrew
3. install following packages by homebrew

brew install cmake
brew install xquartz
brew install gtk+3 boost

4. initiate a virtual environment by conda

cd python_env
./restore_conda_env.sh

1. Web access

http://127.0.0.1:5000

2. Command line

• active conda profile

conda activate face37

'face37' can be customized in python_env/conda_env.conf

• prepare dataset for training

/path/to/dataset
/path/to/dataset/person1/picture1.jpg
/path/to/dataset/person1/picture2.jpg
/path/to/dataset/person2/picture1.jpg
/path/to/dataset/person2/picture2.jpg
/path/to/dataset/person3/picture1.jpg
/path/to/dataset/person3/picture2.jpg

• train a model

python encode_faces.py --dataset /path/to/dataset --encodings model.pickle

• recognize faces from a picture

python recognize_faces_image.py --encodings model.pickle --image one.jpg --display 0