Setup a simple face recognition

We use face_recognition to setup on the Ubuntu server.

Face alignment approach, built using dlib’s state-of-the-art face recognition built with deep learning. The model has an accuracy of 99.38% on the Labeled Faces in the Wild benchmark.

https://github.com/ageitgey/face_recognition

Installation

Ubuntu 18.04.2

Install ssh

sudo apt install openssh-server -y

Development

sudo apt install git python python-pip python-dev python-virtualenv build-essential cmake pkg-config libx11-dev libatlas-base-dev libgtk-3-dev libboost-python-dev virtualenv -y

mkdir -p ~/workspace/python
nano ~/.bash_aliases

Config env path. Declare snip code below to ~/.bash_aliases

#Python env
export ML3_ROOT="~/workspace/python/ml3"
alias ml3="$ML3_ROOT/bin/python3"
alias ml3.setup="virtualenv --system-site-packages -p python3 $ML3_ROOT && ml3.active"
alias ml3.destroy="rm -rf $ML3_ROOT"
alias ml3.active="source $ML3_ROOT/bin/activate"

Active ~/.bash_aliases and setup a python environment

source ~/.bash_aliases

ml3.setup

Python init

# Active python environment
ml3.active

# Install python libs
pip install flask numpy scipy matplotlib pandas statsmodels scikit-learn IPython seaborn nltk plotly cufflinks lightgbm yellowbrick scikit-image dlib opencv-python
pip install --upgrade tensorflow
pip install face_recognition

Run face_recognition example

cd ~/workspace/
git clone https://github.com/ageitgey/face_recognition.git
cd face_recognition

Test with face_recognition cli

# on macOS
~/.local/bin/face_recognition --tolerance 0.54 --show-distance true ./examples/knn_examples/train/alex_lacamoire/ ./examples/knn_examples/test/ | cut -d ',' -f2

# on Ubuntu
face_recognition --tolerance 0.54 --show-distance true ./examples/knn_examples/train/alex_lacamoire/ ./examples/knn_examples/test/ | cut -d ',' -f2

Now create your own server to use face_recognition

Example index.py

# This is a _very simple_ example of a web service that recognizes faces in uploaded images.
# - Upload an image file with name for register face
# - Upload an image file and it will check if the image contains a registered picture.
# The result is returned as json. For example:
#
# $ curl -XPOST -F "file=@obama2.jpg" http://127.0.0.1:5001
#
# Returns:
#
# {
#  "face_found": false,
#  "person": null
# }
#
# This example is based on the Flask file upload example: http://flask.pocoo.org/docs/0.12/patterns/fileuploads/

# NOTE: This example requires flask to be installed! You can install it with pip:
# $ pip3 install flask

import json
import face_recognition
from flask import Flask, jsonify, request, redirect

class Person:
    def __init__(self, name, id, encodings):
        self.name = name
        self.id = id
        self.encodings = encodings

# Alow upload file extensions
ALLOWED_EXTENSIONS = {'png', 'jpg', 'jpeg', 'gif'}
app = Flask(__name__)
# Max content file size 1MB
app.config['MAX_CONTENT_LENGTH'] = 1 * 1024 * 1024

# Store all know face with Person info
person_list = []

def allowed_file(filename):
    return '.' in filename and \
           filename.rsplit('.', 1)[1].lower() in ALLOWED_EXTENSIONS


@app.route('/', methods=['GET'])
def nhancv_face_id():
    return '''
    <!doctype html>
    <title>nhancv.com Face ID</title>
    <h1>For register</h1>
    <form action="/register" target="_blank">
         <button type="submit">Click me</button>
    </form>

    <br>
    <hr>

    <h1>For verify</h1>
    <form action="/verify" target="_blank">
         <button type="submit">Click me</button>
    </form>
    '''

@app.route('/verify', methods=['GET', 'POST'])
def verify_person():
    # Check if a valid image file was uploaded
    if request.method == 'POST':
        if 'file' not in request.files:
            flash('No file part')
            return redirect(request.url)

        file = request.files['file']

        if file.filename == '':
            flash('No file selected for uploading')
            return redirect(request.url)

        if file and allowed_file(file.filename):
            # The image file seems valid! Detect faces and return the result.
            return detect_faces_in_image(file)

    # If no valid image file was uploaded, show the file upload form:
    return '''
    <!doctype html>
    <title>nhancv.com Face ID</title>
    <h1>Upload a picture to verify</h1>
    <form method="POST" url="/verify" enctype="multipart/form-data">
      Photo: <input type="file" name="file"><br>
      <input type="submit" value="Verify">
    </form>
    '''

@app.route('/register', methods=['GET', 'POST'])
def register_person():
    # Check if a valid image file was uploaded
    if request.method == 'POST':
        if 'file' not in request.files:
            return redirect(request.url)

        name = request.form.get('name')
        file = request.files['file']

        if file.filename == '':
            return redirect(request.url)

        if file and allowed_file(file.filename):
            # The image file seems valid! Detect faces and return the result.
            return sample_creating_in_image(name, file)

    # If no valid image file was uploaded, show the file upload form:
    return '''
    <!doctype html>
    <title>Register person</title>
    <h1>Upload a picture to register</h1>
    <form method="POST" url="/register" enctype="multipart/form-data">
      Name: <input type="text" name="name"><br>
      Photo: <input type="file" name="file"><br>
      <input type="submit" value="Register">
    </form>
    '''

def sample_creating_in_image(name, file_stream):
    # Load the uploaded image file
    img = face_recognition.load_image_file(file_stream)
    # Get face encodings for any faces in the uploaded image
    unknown_face_encodings = face_recognition.face_encodings(img)
    
    msg = "Face does not found"
    face_found = False

    if len(unknown_face_encodings) > 0:
        face_found = True
        msg = "Successful"
        # Get input data
        id = len(person_list)
        if not name:
            name = f"Person {id}"
        # Add to array
        person_list.append(Person(name, id, unknown_face_encodings[0]))
        print('Sample person %s:%s created' %(name, id))
    else:
        print (msg)
    # Return the result as json
    result = {
        "error": not face_found,
        "message": msg
    }
    return jsonify(result)


def detect_faces_in_image(file_stream):
    # Extract encodings array
    face_encoding_array = list(map(lambda x: x.encodings, person_list))
    # Load the uploaded image file
    img = face_recognition.load_image_file(file_stream)
    # Get face encodings for any faces in the uploaded image
    unknown_face_encodings = face_recognition.face_encodings(img)

    face_found = False
    person = None
    message = f"Total {len(face_encoding_array)}"

    if len(unknown_face_encodings) > 0:
        # See if the first face in the uploaded image matches the known face
        # face_recognition.api.compare_faces(known_face_encodings, face_encoding_to_check, tolerance=0.6)
        # tolerance – How much distance between faces to consider it a match. Lower is more strict. 0.6 is typical best performance.
        match_results = face_recognition.compare_faces(face_encoding_array, unknown_face_encodings[0], 0.3)
        for i in range(len(match_results)):
            if match_results[i]:
                # person = person_list[i].toJSON()
                person = {
                    "name": person_list[i].name,
                    "id": person_list[i].id
                }
                face_found = True
                message = message + f" - Found: {person_list[i].name}:{person_list[i].id}"
                break;

    print('Status %s. %s' %(face_found, message))
    # Return the result as json
    result = {
        "face_found": face_found,
        "person": person
    }
    return jsonify(result)

if __name__ == "__main__":
    app.run(host='0.0.0.0', port=8888, debug=True)

Start server

python index.py

# Access to localhost:8888

Use on client

register face api
verify face api

For mobile, you can use dlib or firebase vision to detect face rect and crop it, then send it to server to register and verify later.

Clip

Leave a Reply

Your email address will not be published.Required fields are marked *