Face Recognition

Qualeams

Kuhusu programu hii

Face Recognition inaweza kutumika kama mtihani mfumo kwa ajili ya mbinu kadhaa uso utambuzi ikiwa ni pamoja na Neural Networks na TensorFlow na Caffe.

Inajumuisha zifuatazo preprocessing algorithms:
- Grayscale
- Crop
- Eye Mpangilio
- Gamma Masahihisho
- Tofauti ya Gaussians
- Canny-Filter
- Local Binary Pattern
- Histogramm Equalization (inaweza tu kutumika kama kijivujivu hutumiwa pia)
- Resize

Unaweza kuchagua kutoka zifuatazo kipengele uchimbaji na uainishaji mbinu:
- Eigenfaces kwa Karibu Jirani
- Image reshaping na Support Vector Machine
- TensorFlow kwa SVM au KNN
- Caffe kwa SVM au KNN

mwongozo inaweza kupatikana hapa https://github.com/Qualeams/Android-Face-Recognition-with-Deep-Learning/blob/master/USER%20MANUAL.md

Kwa sasa tu armeabi-v7a vifaa na kwenda juu ni mkono.

Kwa matumizi bora katika hali ya utambuzi mzunguko kifaa kushoto.
_______________________________________________________________

TensorFlow:

Kama unataka kutumia Tensorflow Inception5h mfano, kushusha ni kutoka hapa:
https://storage.googleapis.com/download.tensorflow.org/models/inception5h.zip

Kisha nakala faili "tensorflow_inception_graph.pb" na "/ sdcard / Picha / facerecognition / data / TensorFlow"

Matumizi mipangilio hii default kwa ajili ya kuanza:
Idadi ya madarasa: 1001 (si husika kama hatuwezi kutumia safu ya mwisho)
Ingiza ukubwa: 224
Image wastani: 128
Pato ukubwa: 1024
Ingiza safu: pembejeo
Pato safu: avgpool0
Model ya faili: tensorflow_inception_graph.pb
-------------------------------------------------- -------------------------------------------------- -----
Kama unataka kutumia VGG Face descriptor mfano, kushusha ni kutoka hapa:
https://www.dropbox.com/s/51wi2la5e034wfv/vgg_faces.pb?dl=0

Tahadhari: Hii mfano anaendesha tu kwenye vifaa kwa angalau 3 GB au RAM.

Kisha nakala faili "vgg_faces.pb" na "/ sdcard / Picha / facerecognition / data / TensorFlow"

Matumizi mipangilio hii default kwa ajili ya kuanza:
Idadi ya madarasa: 1000 (si muhimu kama sisi wala kutumia safu ya mwisho)
Ingiza ukubwa: 224
Image wastani: 128
Pato ukubwa: 4096
Ingiza safu: Kishika
Pato safu: fc7 / fc7
Model ya faili: vgg_faces.pb
_______________________________________________________________

Caffe:

Kama unataka kutumia VGG Face descriptor mfano, kushusha ni kutoka hapa:
http://www.robots.ox.ac.uk/~vgg/software/vgg_face/src/vgg_face_caffe.tar.gz

Tahadhari: Hii mfano anaendesha tu kwenye vifaa kwa angalau 3 GB au RAM.

Kisha nakala faili "VGG_FACE_deploy.prototxt" na "VGG_FACE.caffemodel" na "/ sdcard / Picha / facerecognition / data / caffe"

Matumizi mipangilio hii default kwa ajili ya kuanza:
maadili Mean: 104, 117, 123
Pato safu: fc7
Model ya faili: VGG_FACE_deploy.prototxt
Uzito faili: VGG_FACE.caffemodel

_______________________________________________________________

files leseni inaweza kupatikana hapa https://github.com/Qualeams/Android-Face-Recognition-with-Deep-Learning/blob/master/LICENSE.txt na hapa https://github.com/Qualeams/Android- ana-Recognition-na-Deep-Learning / kibonge / bwana / NOTICE.txt

Vipengele vipya

- Switch from building Tensorflow from source to using the Jcenter library
- Included optimized_facenet model and changed default settings to use TensorFlow by default