This repo is the face recognition system, which consists of the face detector, the face extractor and the classifier. The face detector is utilized to localize the face in a image, the face extractor is applied to encode the detected face image into the 512-dim feature vectors, and the classifier is employed to classify the feature vector.
This repo consists of three different modules, which is described as below.
|- backbone The network structure of feature extractor
|- dataset The class for loading face images
|- loss The loss function for training the feature extractor
|- preprocess The feature extractor
|- util The util functions for visualize
|- video The demo of the face recognition systems
How to train the feature extractor?
conda activate env-name
python train.py --train_root=datasets\CASIA-WebFace-112x112 --train_file_list=datasets\CASIA-WebFace-112x112.list --lfw_test_root= datasets\lfw-112x112 --lfw_file_list=datasets\pairs.txtHow to fine-tuning the face recognition system?
conda activate env-name
python finetuning.py --train_root=Dataset\Human_Face_Dataset\facebank-112x112 --train_file_list=Dataset\Human_Face_Dataset\facebank-112x112.list