Introduction

• This project refered tf-insightface
• Tensorflow model can be downloaded here

Feature

• v1
  • in support of insightface model
• v2
  • in support of insightface & facenet model
  • add face alignment with higher accuracy, mtcnn output can be linked to tf-insightface directory.

How to build it

• cd thirdparty
• wget https://github.com/opencv/opencv/archive/2.4.13.5.zip
• bash opencv.sh
• cd ..
• mkdir build
• cd build
• cmake ..
• make

How to run it

• put tf model in models directory
• ./bin/demo

Performance

Compared with python verison tf-insightface consuming 118ms, c++ version tf-insightface consumes 87ms, 30ms less on my 8 core cpu machine.