abars / FaceSearchVGG16

Face Search using VGG16 feature value

Face Search VGG16

Implement Face Search using VGG16 feature value

(image from adience_benchmark)

Requirement

Keras + TensorFlow (Feature Extract)

Caffe (Face Detection)

Python 2.7 (Backend)

Perl (Create Dataset)

Preparation

Download Face Detection Pretrained Model

Converted from https://github.com/dannyblueliu/YOLO-version-2-Face-detection

http://www.abars.biz/keras/face.prototxt

http://www.abars.biz/keras/face.caffemodel

Download face.prototxt and face.caffemodel and put in the pretrain folder.

Demo

First, please capture some faces.

Captured face placed in runtime/faces and runtime/feature

python face_search.py capture

Here is a run face seach.

python face_search.py search

Here is a run face seach from file.

python face_search.py search local/faces/landmark_aligned_face.2281.9426695459_9e8b347604_o.jpg

Dataset

Use your own datset

Put face images in local/faces folder.

Here is a extract feature value from local/faces to local/feature.

python feature_extract.py

Use adience benchmark dataset

Download AdienceBenchmarkOfUnfilteredFacesForGenderAndAgeClassification dataset (agegender folder) and put in the dataset folder.

https://www.openu.ac.il/home/hassner/Adience/data.html#agegender

Here is a create local/faces.

perl adience_benchmark_to_faces.pl

Here is a extract feature value from local/faces to local/feature.

python feature_extract.py

Related Work

https://github.com/blan4/KawaiiSearch

https://github.com/matsui528/sis