Skip to content

LeTouristeDeLECAM/AI_Identification_Face_Recognition

Repository files navigation

AI_Identification_Face_Recognition

This project is taking part in larger school project where we implement a photo-gallery identification This have to interpret as a peace of proof

Instructions: https://kenn7.github.io/AIproject/project/

Final project :

Research about the implementation of face recognition

Path : Research/

LBPH

Local Binary Patterns Histograms (LBPH)

YOLO pretrained model

Transfert learning YOLO/ Training YOLO model

Path : Transfert_learning_YOLO/

Description : The goal is to detect if a person is wearing glasses or not. To proceed we've used at first a detection solution but unfortunatly the copped image of the face discard the back ground and do not permit to generalize the detection. So we've used a classification solution.

Data set :

The Dataset is MeGlass, all the face images are selected and cleaned from MegaFace. We used the version composed of copped images 120x120.

Ref : https://github.com/cleardusk/MeGlass

Some processing on the dataset is needed to be done to be able to use it with YOLO. The balacing of the dataset is crucial to avoid bias. The dataset is splited in 2 parts :

  • 80% for training
  • 20% for validation

format :

Classification

The best model is train2 model with 10 epochs and a batch size of 640. The dataset is composed of 1250 faces divided in 2:

  • Test set : Eyeglasses = 174, No Eyeglasses = 126
  • Train set : Eyeglasses = 537, No Eyeglasses = 663

Confusion Matrix

Virtual environment

start virtual environment in terminal

venvFaceRecognition/Scripts/activate

https://www.youtube.com/watch?v=KxvKCSwlUv8

Install requirements

pip install -r requirements.txt

Source :

https://iq.opengenus.org/lbph-algorithm-for-face-recognition/

https://www.freecodecamp.org/news/how-to-detect-objects-in-images-using-yolov8/

@article{guo2018face, title={Face Synthesis for Eyeglass-Robust Face Recognition}, author={Guo, Jianzhu and Zhu, Xiangyu and Lei, Zhen and Li, Stan Z}, journal={arXiv preprint arXiv:1806.01196}, year={2018} }

About

This project is taking part in larger school project where we implement a photo-gallery identification

Resources

Stars

0 stars

Watchers

1 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors