Welcome! My name is Davor and this is my thesis about Face Recognition.

Let's see what can Deep Learning achieve!

---

Tech stack

---

Manual setup

sudo apt-get update
sudo apt-get -y install build-essential

conda create --name exp tensorflow-gpu python=3.7
conda init bash
conda activate exp
conda install -c conda-forge opencv=4.4.0
conda install -c anaconda psycopg2
conda install -c conda-forge pyfiglet
conda install -c conda-forge mtcnn
conda install -c anaconda pillow
conda install -c anaconda flask
conda install -c anaconda flask-cors
pip3 install Cython
pip3 install n2
pip3 install googledrivedownloader

git clone https://github.com/DavorJordacevic/DeepLearningFaceRecognitionSystem
cd DeepLearningFaceRecognitionSystem

wget https://raw.githubusercontent.com/opencv/opencv_extra/master/testdata/dnn/opencv_face_detector.pbtxt
wget https://github.com/spmallick/learnopencv/raw/master/AgeGender/opencv_face_detector_uint8.pb

git clone https://github.com/DavorJordacevic/ArcFace-TF2.x
mv ArcFace-TF2.x arcface_tf2

python downloadCheckpoints.py
rm -rf arc_mbv2.zip
mkdir arcface_tf2/checkpoints
mv arc_mbv2 arcface_tf2/checkpoints/arc_mbv2

git clone https://github.com/DavorJordacevic/retinaface-tf2
mv retinaface-tf2 retinaface_tf2
rm -rf retinaface_mbv2.zip
mkdir retinaface_tf2/checkpoints
mv retinaface_mbv2 retinaface_tf2/checkpoints/retinaface_mbv2

Automatic setup

$ bash frapp.sh

Database model

Frontend

Check: Frontend

Complete system setup (FRAPP, Database, Frontend, ArcFace, RetinaFace)

Check: FRAPP Setup

Face recognition process

---

Start the service

$ ./start.sh

Stop the service

$ ./stop.sh

TODO

• Fix face anti-spoofing
• Docker
• Add multithreading
• Code refactor
• Swagger
• Speed up

---

License

References

peteryuX/arcface-tf2

peteryuX/retinaface-tf2

kakao/N2