Implement Face Search using VGG16 feature value
(image from adience_benchmark)Keras + TensorFlow (Feature Extract)
Caffe (Face Detection)
Python 2.7 (Backend)
Perl (Create Dataset)
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.
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
Put face images in local/faces folder.
Here is a extract feature value from local/faces to local/feature.
python feature_extract.py
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