rebeccawmx / face-recogniton-demo

Face Recognition Demo using dlib and flask

FaceRecognitionDemo using dlib and flask

dlib is a machine learning library used to detect and recognize the faces

flask is a micro framework to create web page using python

Instructions

Open cmd prompt, navigate to the directory

Create conda environment with necessary packages using environment.yml file

conda env create -f environment.yml

Activate the environment

conda activate ./environment

Run app.py

python app.py

Open localhost:5000 in the broswer

Tutorial followed

• One
• Two

Face recognition can be done in following stages

• Face Detection
• Analyze Facial Feautures
• Compare against known faces
• Make a prediction

Face Detection

Face detection is done using Histogram of Oriented Gradients (HOG).In dlib shape_predictor_68_face_landmarks model can be used to get the 68 landmarks of the face.Related code can be found in face_landmark.py file

Facial Features

128 facial measurements (Face embeddings) are genereated by using a pretrained model dlib_face_recognition_resnet_model_v1. Deep learning is used to train the model based on the paper FaceNet

Comparing the faces

These 128 measurements can be used to recognize the face from previously known faces. A Classifier like SVM can also be trained on these facial embeddings

Packages Used

• dlib

conda install -c conda-forge dlib

• Flask

conda install Flask

• Scikit-image

conda install scikit-image