Real-time face recognition with mtcnn, kcf tracker and arcface (insightface)

This is an inference implementation of real time face recogtion, wholely written in C++ with TVM inference runtine. Using mtcnn for facedetection, kcf tracker and insightface model.

MTCNN

[ZHANG2016] Zhang, K., Zhang, Z., Li, Z., and Qiao, Y. (2016). Joint face detection and alignment using multitask cascaded convolutional networks. IEEE Signal Processing Letters, 23(10):1499–1503.

https://kpzhang93.github.io/MTCNN_face_detection_alignment/paper/spl.pdf

Insightface

Deng, Jiankang and Guo, Jia and Niannan, Xue and Zafeiriou, Stefanos. ArcFace: Additive Angular Margin Loss for Deep Face Recognition. CVPR 2019.

ArXiv tecnical report

https://github.com/deepinsight/insightface

Requirements

• OpenCV 3.4+
• Boost FileSystem (1.58+)
• CMake 3.2+
• Clang 6.0+
• TVM runtime

You can folowing this doc file or official doc to install TVM runtime

Build with flowing command

Change two line cmake config in facerecogtion model to your TVM source that you downloaded:

• set (DMLC_INCLUDE "/home/thongpb/works/tvm/3rdparty/dmlc-core/include")
• set (DLPACK_INC "/home/thongpb/works/tvm/3rdparty/dlpack/include")

Compile:

cd realtime-facereccpp
mkdir build
cd build
cmake ..
make -j4

Sample command

You can modify parameters in params.xml

./facerecognition/facerecognition ../data/params.xml

Acknowledgments

The MTCNN implementation take from https://github.com/golunovas/mtcnn-cpp

The model used for facial feature extraction came from insightface MODEL_ZOO

The mtcnn model files are taken from https://github.com/kpzhang93/MTCNN_face_detection_alignment/tree/master/code