awesome_face_antispoofing

• This is a single shot face anti-spoofing project.

• The deep learning framework is Pytorch. Python3.5 is used.

  Installation

• sudo -s

• sh install_requirements.sh

• exit

  Face landmarks

• face_alignment is used for landmarks extraction. Page face_alignment. Thanks to them.

  Landmarks extraction scripts

• cd detlandmark&&python3 detlandmark_imgs.py NUAA_raw_dir

  Data

• I have upload data and detected landmarks into GOOGLE DRIVE-raw.tar.gz

• I have upload data and detected landmarks into Baidu DRIVE-raw.tar.gz

• You can change corresponding directory and filename in config.py

• For example train_filelists=[ ['raw/ClientRaw','raw/client_train_raw.txt',GENUINE], ['raw/ImposterRaw','imposter_train_raw.txt',ATTACK] ] test_filelists=[ ['raw/ClientRaw','raw/client_test_raw.txt',GENUINE], ['raw/ImposterRaw','raw/imposter_test_raw.txt',ATTACK] ]

  Method

• Our method is straightforward. Small patched containing a face is cropped with corresponding landmarks. A binary classification network is used to distinguish the attack patches.

  Training

• First, edit file config.py, choose the target network and proper batch_size.

• Then, in terminal command: make clean&&make&&python3 main.py train

  Inference

• In terminal command: python3 inference.py inference --images='detlandmark/inference_images/*/*.jpg'

• The inference report is result/inference.txt, you can check it in commad: cat result/inference.txt

  Visualize Dataset

• We have fixed the bug of choice wrong face in multiple detected faces with standard of coordinates.

• To visualize cropped faces in dataset. Please run command: python3 vis_cropface.py visualize

• All faces will be shown in data/showcropface_train.jpg and data/showcropface_val.jpg

• The training data are shown here. Training Validation

Experiments

• Experiments results on NUAA[1] Image input size is as same as the imagenet.
• State-of-the-art networks are used, e.g. VGG[2], ResNet[3], DenseNet[4], Inception[5], Xception[6], DetNet[7]

Reference

• [1]Tan X, Li Y, Liu J, et al. Face liveness detection from a single image with sparse low rank bilinear discriminative model[C]// European Conference on Computer Vision. Springer-Verlag, 2010:504-517.
• [2]Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. arXiv preprint arXiv:1409.1556, 2014.
• [3]He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 770-778.
• [4]Huang G, Liu Z, Van Der Maaten L, et al. Densely connected convolutional networks[C]//CVPR. 2017, 1(2): 3.
• [5]Szegedy C, Vanhoucke V, Ioffe S, et al. Rethinking the inception architecture for computer vision[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2016: 2818-2826.
• [6]Chollet F. Xception: Deep learning with depthwise separable convolutions[J]. arXiv preprint, 2017: 1610.02357.
• [7]Li Z, Peng C, Yu G, et al. DetNet: A Backbone network for Object Detection[J]. arXiv preprint arXiv:1804.06215, 2018.