xhu248 / arcface_noonan

Arcface_for_Noonan_detection

Pytorch1.7.1 codes for arcface

---

1. Intro

• This repo is the source code of paper "Automated Facial Recognition for Noonan Syndrome using Deep Convolutional Neural Network with Additive Angular Margin Loss" (under review)

---

2. Pretrained Models & Performance

The pretrained models are from: https://github.com/TreB1eN/InsightFace_Pytorch

Mobilefacenet @ BaiduNetDisk, Mobilefacenet @ OneDrive

LFW(%)	CFP-FF(%)	CFP-FP(%)	AgeDB-30(%)	calfw(%)	cplfw(%)	vgg2_fp(%)
0.9918	0.9891	0.8986	0.9347	0.9402	0.866	0.9100

3. How to use

• download codes

  git clone [e-mail]:lucasforever24/arcface_noonan.git
  

3.1 Data Preparation

3.1.1 Prepare Facebank

Place the face images your want to detect in the data/facebank/noonan folder, and guarantee it have a structure like following:

data/facebank/noonan
        ---> noonan1/
            ---> noonan1_1.jpg
        ---> normal1/
            ---> normal2_1.jpg
        ---> others1/
            ---> others1_1.jpg
           ---> others1_2.jpg

Noonan stands for Noonan syndrome patients, normal stands for healthy children, and others refer to children with other syndromes.

3.1.2 download the pretrained model to work_space/model

If more than 1 image appears in one folder, an average embedding will be calculated

3.2 run:

• bash run.sh
• or python fold_cur_trans.py -ds {the folder name in facebank}

4. References

• This repo is mainly inspired by https://github.com/TreB1eN/InsightFace_Pytorch