This is a test script for the VGG_face deep model. Please feel free to use it. We explicitly have not subtracted the mean face, which is verified to be better on YouTube Face dataset. It simply compares the correlation between two deeply learned features corresponding with two testing facial images needed to be verified.
Face recognition seems rising again. Google, Facebook and SenseTime have almost solved it on benchmarks such as Labelled Face in the Wild (LFW). Their face recognizer are known as FaceNet, DeepFace and FaceID, respectively. Andrew Zisserman's group from the University of Oxford released their deep face model called VGG Face Descriptor, which is claimed to be trained on millions of face images. Today I take it and give it a shot on the task of face verification - determining whether a pair of two face images are from the same person or not. I simply compare the correlation (or cosine-similarity) between two deeply learned features corresponding with the two testing facial images that needed to be verified. Below are some results on LFW. Does the similarity (the third column) really tell you if the left person and the right one are the same person?
The program uses CPU in default. If you want to try GPU mode, please simply comment the first line in src/Classifier.cpp
VGG pre-trained model and the Caffe protocol file can be downloaded from http://www.robots.ox.ac.uk/~vgg/software/vgg_face/src/vgg_face_caffe.tar.gz
Features are outputs of FC8 so they are 4096-dim.
========== Settings ============
Depending on the operating system, you may set link libraries and include headfiles in the following way.
/usr/lib64/libboost_system.so
/usr/local/lib/libopencv_highgui.so
/usr/local/lib/libopencv_core.so
/usr/local/lib/libopencv_imgproc.so
/usr/lib64/libglog.so
/usr/local/lib/libcaffe.so
/usr/local/lib/libopencv_imgcodecs.so
Linker:
/usr/local/lib
/usr/lib
Compiler:
/usr/local/include
http://stackoverflow.com/questions/33828582/vgg-face-descriptor-in-python-with-caffe
Most people argued if subtracting the mean face is a necessity. From my experiments, I would say no. You do not need to subtract a fake mean to the input images.
I am more than happy to know if this program is of use to your work. Please cite the following paper.
@inproceedings{xiang16pose,
author = {Xiang, Xiang and Tran, Trac D},
Title = {Pose-Selective Max Pooling for Measuring Similarity},
booktitle = {VAAM/FFER@ICPR 2016},
Year = {2016}
}