Introduction

A tool for face detection, alignment and verification

How to Run

To run this tool, you need to:

• Install Caffe: https://github.com/BVLC/caffe
• Change the CAFFE_PATH in "utils/init_path.py"
• Download models from Dropbox or BaiduWangPan, and put them in "model/"

Then you can run "demo.py" to see the result. If everything is right, you will see two faces are detected and their names (Rose and Jack) are labled.

Reference

The face detection and alignment algorithm named MTCNN is from this paper:

• Kaipeng Zhang, Zhanpeng Zhang, Zhifeng Li, Yu Qiao , " Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Networks," IEEE Signal Processing Letter

And we also get some code from https://github.com/DuinoDu/mtcnn, which is a python version of MTCNN.

The verification algorithm is from MMLab of CUHK.