FaceRecognitioncanbeusedasatestframeworkforseveralfacerecognitionmethodsincludingtheNeuralNetworkswithTensorFlowandCaffe.
Itincludesfollowingpreprocessingalgorithms:
-Grayscale
-Crop
-EyeAlignment
-GammaCorrection
-DifferenceofGaussians
-Canny-Filter
-LocalBinaryPattern
-HistogrammEqualization(canonlybeusedifgrayscaleisusedtoo)
-Resize
Youcanchoosefromthefollowingfeatureextractionandclassificationmethods:
-EigenfaceswithNearestNeighbour
-ImageReshapingwithSupportVectorMachine
-TensorFlowwithSVMorKNN
-CaffewithSVMorKNN
Themanualcanbefoundherehttps://github.com/Qualeams/Android-Face-Recognition-with-Deep-Learning/blob/master/USER%20MANUAL.md
Atthemomentonlyarmeabi-v7adevicesandupwardsaresupported.
Forbestexperienceinrecognitionmoderotatethedevicetoleft.
Forbestperformanceuse"ImageReshapingwithSupportVectorMachine"(0.5s/image).
Forbestaccuracyusethe"VGGFaceDescriptor"model(theperformanceisverybadthough-6.5s/image)
_______________________________________________________________
TensorFlow:
IfyouwanttousetheTensorflowInception5hmodel,downloaditfromhere:
https://storage.googleapis.com/download.tensorflow.org/models/inception5h.zip
Thencopythefile"tensorflow_inception_graph.pb"to"/sdcard/Pictures/facerecognition/data/TensorFlow"
Usethesedefaultsettingsforastart:
Numberofclasses:1001(notrelevantaswedon'tusethelastlayer)
InputSize:224
Imagemean:128
Outputsize:1024
Inputlayer:input
Outputlayer:avgpool0
Modelfile:tensorflow_inception_graph.pb
---------------------------------------------------------------------------------------------------------
IfyouwanttousetheVGGFaceDescriptormodel,downloaditfromhere:
https://www.dropbox.com/s/51wi2la5e034wfv/vgg_faces.pb?dl=0
Caution:Thismodelrunsonlyondeviceswithatleast3GBorRAM.
Thencopythefile"vgg_faces.pb"to"/sdcard/Pictures/facerecognition/data/TensorFlow"
Usethesedefaultsettingsforastart:
Numberofclasses:1000(notrelevantaswedon'tusethelastlayer)
InputSize:224
Imagemean:128
Outputsize:4096
Inputlayer:Placeholder
Outputlayer:fc7/fc7
Modelfile:vgg_faces.pb
_______________________________________________________________
Caffe:
IfyouwanttousetheVGGFaceDescriptormodel,downloaditfromhere:
http://www.robots.ox.ac.uk/~vgg/software/vgg_face/src/vgg_face_caffe.tar.gz
Caution:Thismodelrunsonlyondeviceswithatleast3GBorRAM.
Thencopythefiles"VGG_FACE_deploy.prototxt"and"VGG_FACE.caffemodel"to"/sdcard/Pictures/facerecognition/data/caffe"
Usethesedefaultsettingsforastart:
Meanvalues:104,117,123
Outputlayer:fc7
Modelfile:VGG_FACE_deploy.prototxt
Weightsfile:VGG_FACE.caffemodel
_______________________________________________________________
Thelicensefilescanbefoundherehttps://github.com/Qualeams/Android-Face-Recognition-with-Deep-Learning/blob/master/LICENSE.txtandherehttps://github.com/Qualeams/Android-Face-Recognition-with-Deep-Learning/blob/master/NOTICE.txt