An open source face recognition engine.
Support for Tensorflow-2.0.0+ @[branch tensorflow2.0.0+]
Let us see something about this project now.
This project is an open source project about face recognition. In the project, we implemente a face recognition engine with one-shot training.
In this project, we implemente some CNNs (VGGNet, VIPL face net, ResNet, XCEPTION, et al) to recognize face image.
Here, we use AM-Softmax loss as the cost function rather than triple loss or other metric learning loss functions because AM-Softmax has less training time but accuracy is still good.
This project is only a demo. In order to see the experimental results, we trained a model with small data. We use data augmentation in this project, so that we can get a robust model. If you want to use this project in your production environment, you should do more.
python3 train.py
python3 predict.py
Python 3.6+
Others:
```
pip3 install -r requirements.txt
```
You can search the following papers in Google Scholar
AM-Softmax
Sphere face
FaceNet
ResNet
Xception
MobileNet v1,v2,v3
VIPL Face net
- https://github.com/xiangrufan/keras-mtcnn
- https://github.com/happynear/AMSoftmax
- https://github.com/Joker316701882/Additive-Margin-Softmax
- https://github.com/hao-qiang/AM-Softmax
- https://github.com/ageitgey/face_recognition
- https://github.com/oarriaga/face_classification
- https://github.com/seetaface/SeetaFaceEngine
- https://github.com/jiankangdeng/handbook
Apache license version 2.0
There are many bugs here, so you could send some pull requests or give some issues for this project. Thank you very much :)
- give train.py arguments: for different training set
- refactor: to optimize code
- etc.