Hugging Face's logo

Spaces:
AshanGimhana
/
Aging_MouthReplace
Paused

App Files Community
Aging_MouthReplace / dlibs /docs /dnn_face_recognition_ex.cpp.html
AshanGimhana's picture
AshanGimhana
Upload folder using huggingface_hub
9375c9a verified
raw
history blame contribute delete
31.9 kB
<html><!-- Created using the cpp_pretty_printer from the dlib C++ library. See http://dlib.net for updates. --><head><title>dlib C++ Library - dnn_face_recognition_ex.cpp</title></head><body bgcolor='white'><pre>
<font color='#009900'>// The contents of this file are in the public domain. See LICENSE_FOR_EXAMPLE_PROGRAMS.txt
</font><font color='#009900'>/*
This is an example illustrating the use of the deep learning tools from the dlib C++
Library. In it, we will show how to do face recognition. This example uses the
pretrained dlib_face_recognition_resnet_model_v1 model which is freely available from
the dlib web site. This model has a 99.38% accuracy on the standard LFW face
recognition benchmark, which is comparable to other state-of-the-art methods for face
recognition as of February 2017.
In this example, we will use dlib to do face clustering. Included in the examples
folder is an image, bald_guys.jpg, which contains a bunch of photos of action movie
stars Vin Diesel, The Rock, Jason Statham, and Bruce Willis. We will use dlib to
automatically find their faces in the image and then to automatically determine how
many people there are (4 in this case) as well as which faces belong to each person.
Finally, this example uses a network with the loss_metric loss. Therefore, if you want
to learn how to train your own models, or to get a general introduction to this loss
layer, you should read the <a href="dnn_metric_learning_ex.cpp.html">dnn_metric_learning_ex.cpp</a> and
<a href="dnn_metric_learning_on_images_ex.cpp.html">dnn_metric_learning_on_images_ex.cpp</a> examples.
*/</font>
<font color='#0000FF'>#include</font> <font color='#5555FF'>&lt;</font>dlib<font color='#5555FF'>/</font>dnn.h<font color='#5555FF'>&gt;</font>
<font color='#0000FF'>#include</font> <font color='#5555FF'>&lt;</font>dlib<font color='#5555FF'>/</font>gui_widgets.h<font color='#5555FF'>&gt;</font>
<font color='#0000FF'>#include</font> <font color='#5555FF'>&lt;</font>dlib<font color='#5555FF'>/</font>clustering.h<font color='#5555FF'>&gt;</font>
<font color='#0000FF'>#include</font> <font color='#5555FF'>&lt;</font>dlib<font color='#5555FF'>/</font>string.h<font color='#5555FF'>&gt;</font>
<font color='#0000FF'>#include</font> <font color='#5555FF'>&lt;</font>dlib<font color='#5555FF'>/</font>image_io.h<font color='#5555FF'>&gt;</font>
<font color='#0000FF'>#include</font> <font color='#5555FF'>&lt;</font>dlib<font color='#5555FF'>/</font>image_processing<font color='#5555FF'>/</font>frontal_face_detector.h<font color='#5555FF'>&gt;</font>
<font color='#0000FF'>using</font> <font color='#0000FF'>namespace</font> dlib;
<font color='#0000FF'>using</font> <font color='#0000FF'>namespace</font> std;
<font color='#009900'>// ----------------------------------------------------------------------------------------
</font>
<font color='#009900'>// The next bit of code defines a ResNet network. It's basically copied
</font><font color='#009900'>// and pasted from the <a href="dnn_imagenet_ex.cpp.html">dnn_imagenet_ex.cpp</a> example, except we replaced the loss
</font><font color='#009900'>// layer with loss_metric and made the network somewhat smaller. Go read the introductory
</font><font color='#009900'>// dlib DNN examples to learn what all this stuff means.
</font><font color='#009900'>//
</font><font color='#009900'>// Also, the <a href="dnn_metric_learning_on_images_ex.cpp.html">dnn_metric_learning_on_images_ex.cpp</a> example shows how to train this network.
</font><font color='#009900'>// The dlib_face_recognition_resnet_model_v1 model used by this example was trained using
</font><font color='#009900'>// essentially the code shown in <a href="dnn_metric_learning_on_images_ex.cpp.html">dnn_metric_learning_on_images_ex.cpp</a> except the
</font><font color='#009900'>// mini-batches were made larger (35x15 instead of 5x5), the iterations without progress
</font><font color='#009900'>// was set to 10000, and the training dataset consisted of about 3 million images instead of
</font><font color='#009900'>// 55. Also, the input layer was locked to images of size 150.
</font><font color='#0000FF'>template</font> <font color='#5555FF'>&lt;</font><font color='#0000FF'>template</font> <font color='#5555FF'>&lt;</font><font color='#0000FF'><u>int</u></font>,<font color='#0000FF'>template</font><font color='#5555FF'>&lt;</font><font color='#0000FF'>typename</font><font color='#5555FF'>&gt;</font><font color='#0000FF'>class</font>,<font color='#0000FF'><u>int</u></font>,<font color='#0000FF'>typename</font><font color='#5555FF'>&gt;</font> <font color='#0000FF'>class</font> <b><a name='block'></a>block</b>, <font color='#0000FF'><u>int</u></font> N, <font color='#0000FF'>template</font><font color='#5555FF'>&lt;</font><font color='#0000FF'>typename</font><font color='#5555FF'>&gt;</font><font color='#0000FF'>class</font> <b><a name='BN'></a>BN</b>, <font color='#0000FF'>typename</font> SUBNET<font color='#5555FF'>&gt;</font>
<font color='#0000FF'>using</font> residual <font color='#5555FF'>=</font> add_prev1<font color='#5555FF'>&lt;</font>block<font color='#5555FF'>&lt;</font>N,BN,<font color='#979000'>1</font>,tag1<font color='#5555FF'>&lt;</font>SUBNET<font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font>;
<font color='#0000FF'>template</font> <font color='#5555FF'>&lt;</font><font color='#0000FF'>template</font> <font color='#5555FF'>&lt;</font><font color='#0000FF'><u>int</u></font>,<font color='#0000FF'>template</font><font color='#5555FF'>&lt;</font><font color='#0000FF'>typename</font><font color='#5555FF'>&gt;</font><font color='#0000FF'>class</font>,<font color='#0000FF'><u>int</u></font>,<font color='#0000FF'>typename</font><font color='#5555FF'>&gt;</font> <font color='#0000FF'>class</font> <b><a name='block'></a>block</b>, <font color='#0000FF'><u>int</u></font> N, <font color='#0000FF'>template</font><font color='#5555FF'>&lt;</font><font color='#0000FF'>typename</font><font color='#5555FF'>&gt;</font><font color='#0000FF'>class</font> <b><a name='BN'></a>BN</b>, <font color='#0000FF'>typename</font> SUBNET<font color='#5555FF'>&gt;</font>
<font color='#0000FF'>using</font> residual_down <font color='#5555FF'>=</font> add_prev2<font color='#5555FF'>&lt;</font>avg_pool<font color='#5555FF'>&lt;</font><font color='#979000'>2</font>,<font color='#979000'>2</font>,<font color='#979000'>2</font>,<font color='#979000'>2</font>,skip1<font color='#5555FF'>&lt;</font>tag2<font color='#5555FF'>&lt;</font>block<font color='#5555FF'>&lt;</font>N,BN,<font color='#979000'>2</font>,tag1<font color='#5555FF'>&lt;</font>SUBNET<font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font>;
<font color='#0000FF'>template</font> <font color='#5555FF'>&lt;</font><font color='#0000FF'><u>int</u></font> N, <font color='#0000FF'>template</font> <font color='#5555FF'>&lt;</font><font color='#0000FF'>typename</font><font color='#5555FF'>&gt;</font> <font color='#0000FF'>class</font> <b><a name='BN'></a>BN</b>, <font color='#0000FF'><u>int</u></font> stride, <font color='#0000FF'>typename</font> SUBNET<font color='#5555FF'>&gt;</font>
<font color='#0000FF'>using</font> block <font color='#5555FF'>=</font> BN<font color='#5555FF'>&lt;</font>con<font color='#5555FF'>&lt;</font>N,<font color='#979000'>3</font>,<font color='#979000'>3</font>,<font color='#979000'>1</font>,<font color='#979000'>1</font>,relu<font color='#5555FF'>&lt;</font>BN<font color='#5555FF'>&lt;</font>con<font color='#5555FF'>&lt;</font>N,<font color='#979000'>3</font>,<font color='#979000'>3</font>,stride,stride,SUBNET<font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font>;
<font color='#0000FF'>template</font> <font color='#5555FF'>&lt;</font><font color='#0000FF'><u>int</u></font> N, <font color='#0000FF'>typename</font> SUBNET<font color='#5555FF'>&gt;</font> <font color='#0000FF'>using</font> ares <font color='#5555FF'>=</font> relu<font color='#5555FF'>&lt;</font>residual<font color='#5555FF'>&lt;</font>block,N,affine,SUBNET<font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font>;
<font color='#0000FF'>template</font> <font color='#5555FF'>&lt;</font><font color='#0000FF'><u>int</u></font> N, <font color='#0000FF'>typename</font> SUBNET<font color='#5555FF'>&gt;</font> <font color='#0000FF'>using</font> ares_down <font color='#5555FF'>=</font> relu<font color='#5555FF'>&lt;</font>residual_down<font color='#5555FF'>&lt;</font>block,N,affine,SUBNET<font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font>;
<font color='#0000FF'>template</font> <font color='#5555FF'>&lt;</font><font color='#0000FF'>typename</font> SUBNET<font color='#5555FF'>&gt;</font> <font color='#0000FF'>using</font> alevel0 <font color='#5555FF'>=</font> ares_down<font color='#5555FF'>&lt;</font><font color='#979000'>256</font>,SUBNET<font color='#5555FF'>&gt;</font>;
<font color='#0000FF'>template</font> <font color='#5555FF'>&lt;</font><font color='#0000FF'>typename</font> SUBNET<font color='#5555FF'>&gt;</font> <font color='#0000FF'>using</font> alevel1 <font color='#5555FF'>=</font> ares<font color='#5555FF'>&lt;</font><font color='#979000'>256</font>,ares<font color='#5555FF'>&lt;</font><font color='#979000'>256</font>,ares_down<font color='#5555FF'>&lt;</font><font color='#979000'>256</font>,SUBNET<font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font>;
<font color='#0000FF'>template</font> <font color='#5555FF'>&lt;</font><font color='#0000FF'>typename</font> SUBNET<font color='#5555FF'>&gt;</font> <font color='#0000FF'>using</font> alevel2 <font color='#5555FF'>=</font> ares<font color='#5555FF'>&lt;</font><font color='#979000'>128</font>,ares<font color='#5555FF'>&lt;</font><font color='#979000'>128</font>,ares_down<font color='#5555FF'>&lt;</font><font color='#979000'>128</font>,SUBNET<font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font>;
<font color='#0000FF'>template</font> <font color='#5555FF'>&lt;</font><font color='#0000FF'>typename</font> SUBNET<font color='#5555FF'>&gt;</font> <font color='#0000FF'>using</font> alevel3 <font color='#5555FF'>=</font> ares<font color='#5555FF'>&lt;</font><font color='#979000'>64</font>,ares<font color='#5555FF'>&lt;</font><font color='#979000'>64</font>,ares<font color='#5555FF'>&lt;</font><font color='#979000'>64</font>,ares_down<font color='#5555FF'>&lt;</font><font color='#979000'>64</font>,SUBNET<font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font>;
<font color='#0000FF'>template</font> <font color='#5555FF'>&lt;</font><font color='#0000FF'>typename</font> SUBNET<font color='#5555FF'>&gt;</font> <font color='#0000FF'>using</font> alevel4 <font color='#5555FF'>=</font> ares<font color='#5555FF'>&lt;</font><font color='#979000'>32</font>,ares<font color='#5555FF'>&lt;</font><font color='#979000'>32</font>,ares<font color='#5555FF'>&lt;</font><font color='#979000'>32</font>,SUBNET<font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font>;
<font color='#0000FF'>using</font> anet_type <font color='#5555FF'>=</font> loss_metric<font color='#5555FF'>&lt;</font>fc_no_bias<font color='#5555FF'>&lt;</font><font color='#979000'>128</font>,avg_pool_everything<font color='#5555FF'>&lt;</font>
alevel0<font color='#5555FF'>&lt;</font>
alevel1<font color='#5555FF'>&lt;</font>
alevel2<font color='#5555FF'>&lt;</font>
alevel3<font color='#5555FF'>&lt;</font>
alevel4<font color='#5555FF'>&lt;</font>
max_pool<font color='#5555FF'>&lt;</font><font color='#979000'>3</font>,<font color='#979000'>3</font>,<font color='#979000'>2</font>,<font color='#979000'>2</font>,relu<font color='#5555FF'>&lt;</font>affine<font color='#5555FF'>&lt;</font>con<font color='#5555FF'>&lt;</font><font color='#979000'>32</font>,<font color='#979000'>7</font>,<font color='#979000'>7</font>,<font color='#979000'>2</font>,<font color='#979000'>2</font>,
input_rgb_image_sized<font color='#5555FF'>&lt;</font><font color='#979000'>150</font><font color='#5555FF'>&gt;</font>
<font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font>;
<font color='#009900'>// ----------------------------------------------------------------------------------------
</font>
std::vector<font color='#5555FF'>&lt;</font>matrix<font color='#5555FF'>&lt;</font>rgb_pixel<font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font> <b><a name='jitter_image'></a>jitter_image</b><font face='Lucida Console'>(</font>
<font color='#0000FF'>const</font> matrix<font color='#5555FF'>&lt;</font>rgb_pixel<font color='#5555FF'>&gt;</font><font color='#5555FF'>&amp;</font> img
<font face='Lucida Console'>)</font>;
<font color='#009900'>// ----------------------------------------------------------------------------------------
</font>
<font color='#0000FF'><u>int</u></font> <b><a name='main'></a>main</b><font face='Lucida Console'>(</font><font color='#0000FF'><u>int</u></font> argc, <font color='#0000FF'><u>char</u></font><font color='#5555FF'>*</font><font color='#5555FF'>*</font> argv<font face='Lucida Console'>)</font> <font color='#0000FF'>try</font>
<b>{</b>
<font color='#0000FF'>if</font> <font face='Lucida Console'>(</font>argc <font color='#5555FF'>!</font><font color='#5555FF'>=</font> <font color='#979000'>2</font><font face='Lucida Console'>)</font>
<b>{</b>
cout <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> "<font color='#CC0000'>Run this example by invoking it like this: </font>" <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> endl;
cout <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> "<font color='#CC0000'> ./dnn_face_recognition_ex faces/bald_guys.jpg</font>" <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> endl;
cout <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> endl;
cout <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> "<font color='#CC0000'>You will also need to get the face landmarking model file as well as </font>" <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> endl;
cout <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> "<font color='#CC0000'>the face recognition model file. Download and then decompress these files from: </font>" <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> endl;
cout <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> "<font color='#CC0000'>http://dlib.net/files/shape_predictor_5_face_landmarks.dat.bz2</font>" <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> endl;
cout <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> "<font color='#CC0000'>http://dlib.net/files/dlib_face_recognition_resnet_model_v1.dat.bz2</font>" <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> endl;
cout <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> endl;
<font color='#0000FF'>return</font> <font color='#979000'>1</font>;
<b>}</b>
<font color='#009900'>// The first thing we are going to do is load all our models. First, since we need to
</font> <font color='#009900'>// find faces in the image we will need a face detector:
</font> frontal_face_detector detector <font color='#5555FF'>=</font> <font color='#BB00BB'>get_frontal_face_detector</font><font face='Lucida Console'>(</font><font face='Lucida Console'>)</font>;
<font color='#009900'>// We will also use a face landmarking model to align faces to a standard pose: (see <a href="face_landmark_detection_ex.cpp.html">face_landmark_detection_ex.cpp</a> for an introduction)
</font> shape_predictor sp;
<font color='#BB00BB'>deserialize</font><font face='Lucida Console'>(</font>"<font color='#CC0000'>shape_predictor_5_face_landmarks.dat</font>"<font face='Lucida Console'>)</font> <font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font> sp;
<font color='#009900'>// And finally we load the DNN responsible for face recognition.
</font> anet_type net;
<font color='#BB00BB'>deserialize</font><font face='Lucida Console'>(</font>"<font color='#CC0000'>dlib_face_recognition_resnet_model_v1.dat</font>"<font face='Lucida Console'>)</font> <font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font> net;
matrix<font color='#5555FF'>&lt;</font>rgb_pixel<font color='#5555FF'>&gt;</font> img;
<font color='#BB00BB'>load_image</font><font face='Lucida Console'>(</font>img, argv[<font color='#979000'>1</font>]<font face='Lucida Console'>)</font>;
<font color='#009900'>// Display the raw image on the screen
</font> image_window <font color='#BB00BB'>win</font><font face='Lucida Console'>(</font>img<font face='Lucida Console'>)</font>;
<font color='#009900'>// Run the face detector on the image of our action heroes, and for each face extract a
</font> <font color='#009900'>// copy that has been normalized to 150x150 pixels in size and appropriately rotated
</font> <font color='#009900'>// and centered.
</font> std::vector<font color='#5555FF'>&lt;</font>matrix<font color='#5555FF'>&lt;</font>rgb_pixel<font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font> faces;
<font color='#0000FF'>for</font> <font face='Lucida Console'>(</font><font color='#0000FF'>auto</font> face : <font color='#BB00BB'>detector</font><font face='Lucida Console'>(</font>img<font face='Lucida Console'>)</font><font face='Lucida Console'>)</font>
<b>{</b>
<font color='#0000FF'>auto</font> shape <font color='#5555FF'>=</font> <font color='#BB00BB'>sp</font><font face='Lucida Console'>(</font>img, face<font face='Lucida Console'>)</font>;
matrix<font color='#5555FF'>&lt;</font>rgb_pixel<font color='#5555FF'>&gt;</font> face_chip;
<font color='#BB00BB'>extract_image_chip</font><font face='Lucida Console'>(</font>img, <font color='#BB00BB'>get_face_chip_details</font><font face='Lucida Console'>(</font>shape,<font color='#979000'>150</font>,<font color='#979000'>0.25</font><font face='Lucida Console'>)</font>, face_chip<font face='Lucida Console'>)</font>;
faces.<font color='#BB00BB'>push_back</font><font face='Lucida Console'>(</font><font color='#BB00BB'>move</font><font face='Lucida Console'>(</font>face_chip<font face='Lucida Console'>)</font><font face='Lucida Console'>)</font>;
<font color='#009900'>// Also put some boxes on the faces so we can see that the detector is finding
</font> <font color='#009900'>// them.
</font> win.<font color='#BB00BB'>add_overlay</font><font face='Lucida Console'>(</font>face<font face='Lucida Console'>)</font>;
<b>}</b>
<font color='#0000FF'>if</font> <font face='Lucida Console'>(</font>faces.<font color='#BB00BB'>size</font><font face='Lucida Console'>(</font><font face='Lucida Console'>)</font> <font color='#5555FF'>=</font><font color='#5555FF'>=</font> <font color='#979000'>0</font><font face='Lucida Console'>)</font>
<b>{</b>
cout <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> "<font color='#CC0000'>No faces found in image!</font>" <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> endl;
<font color='#0000FF'>return</font> <font color='#979000'>1</font>;
<b>}</b>
<font color='#009900'>// This call asks the DNN to convert each face image in faces into a 128D vector.
</font> <font color='#009900'>// In this 128D vector space, images from the same person will be close to each other
</font> <font color='#009900'>// but vectors from different people will be far apart. So we can use these vectors to
</font> <font color='#009900'>// identify if a pair of images are from the same person or from different people.
</font> std::vector<font color='#5555FF'>&lt;</font>matrix<font color='#5555FF'>&lt;</font><font color='#0000FF'><u>float</u></font>,<font color='#979000'>0</font>,<font color='#979000'>1</font><font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font> face_descriptors <font color='#5555FF'>=</font> <font color='#BB00BB'>net</font><font face='Lucida Console'>(</font>faces<font face='Lucida Console'>)</font>;
<font color='#009900'>// In particular, one simple thing we can do is face clustering. This next bit of code
</font> <font color='#009900'>// creates a graph of connected faces and then uses the Chinese whispers graph clustering
</font> <font color='#009900'>// algorithm to identify how many people there are and which faces belong to whom.
</font> std::vector<font color='#5555FF'>&lt;</font>sample_pair<font color='#5555FF'>&gt;</font> edges;
<font color='#0000FF'>for</font> <font face='Lucida Console'>(</font><font color='#0000FF'><u>size_t</u></font> i <font color='#5555FF'>=</font> <font color='#979000'>0</font>; i <font color='#5555FF'>&lt;</font> face_descriptors.<font color='#BB00BB'>size</font><font face='Lucida Console'>(</font><font face='Lucida Console'>)</font>; <font color='#5555FF'>+</font><font color='#5555FF'>+</font>i<font face='Lucida Console'>)</font>
<b>{</b>
<font color='#0000FF'>for</font> <font face='Lucida Console'>(</font><font color='#0000FF'><u>size_t</u></font> j <font color='#5555FF'>=</font> i; j <font color='#5555FF'>&lt;</font> face_descriptors.<font color='#BB00BB'>size</font><font face='Lucida Console'>(</font><font face='Lucida Console'>)</font>; <font color='#5555FF'>+</font><font color='#5555FF'>+</font>j<font face='Lucida Console'>)</font>
<b>{</b>
<font color='#009900'>// Faces are connected in the graph if they are close enough. Here we check if
</font> <font color='#009900'>// the distance between two face descriptors is less than 0.6, which is the
</font> <font color='#009900'>// decision threshold the network was trained to use. Although you can
</font> <font color='#009900'>// certainly use any other threshold you find useful.
</font> <font color='#0000FF'>if</font> <font face='Lucida Console'>(</font><font color='#BB00BB'>length</font><font face='Lucida Console'>(</font>face_descriptors[i]<font color='#5555FF'>-</font>face_descriptors[j]<font face='Lucida Console'>)</font> <font color='#5555FF'>&lt;</font> <font color='#979000'>0.6</font><font face='Lucida Console'>)</font>
edges.<font color='#BB00BB'>push_back</font><font face='Lucida Console'>(</font><font color='#BB00BB'>sample_pair</font><font face='Lucida Console'>(</font>i,j<font face='Lucida Console'>)</font><font face='Lucida Console'>)</font>;
<b>}</b>
<b>}</b>
std::vector<font color='#5555FF'>&lt;</font><font color='#0000FF'><u>unsigned</u></font> <font color='#0000FF'><u>long</u></font><font color='#5555FF'>&gt;</font> labels;
<font color='#0000FF'>const</font> <font color='#0000FF'>auto</font> num_clusters <font color='#5555FF'>=</font> <font color='#BB00BB'>chinese_whispers</font><font face='Lucida Console'>(</font>edges, labels<font face='Lucida Console'>)</font>;
<font color='#009900'>// This will correctly indicate that there are 4 people in the image.
</font> cout <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> "<font color='#CC0000'>number of people found in the image: </font>"<font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> num_clusters <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> endl;
<font color='#009900'>// Now let's display the face clustering results on the screen. You will see that it
</font> <font color='#009900'>// correctly grouped all the faces.
</font> std::vector<font color='#5555FF'>&lt;</font>image_window<font color='#5555FF'>&gt;</font> <font color='#BB00BB'>win_clusters</font><font face='Lucida Console'>(</font>num_clusters<font face='Lucida Console'>)</font>;
<font color='#0000FF'>for</font> <font face='Lucida Console'>(</font><font color='#0000FF'><u>size_t</u></font> cluster_id <font color='#5555FF'>=</font> <font color='#979000'>0</font>; cluster_id <font color='#5555FF'>&lt;</font> num_clusters; <font color='#5555FF'>+</font><font color='#5555FF'>+</font>cluster_id<font face='Lucida Console'>)</font>
<b>{</b>
std::vector<font color='#5555FF'>&lt;</font>matrix<font color='#5555FF'>&lt;</font>rgb_pixel<font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font> temp;
<font color='#0000FF'>for</font> <font face='Lucida Console'>(</font><font color='#0000FF'><u>size_t</u></font> j <font color='#5555FF'>=</font> <font color='#979000'>0</font>; j <font color='#5555FF'>&lt;</font> labels.<font color='#BB00BB'>size</font><font face='Lucida Console'>(</font><font face='Lucida Console'>)</font>; <font color='#5555FF'>+</font><font color='#5555FF'>+</font>j<font face='Lucida Console'>)</font>
<b>{</b>
<font color='#0000FF'>if</font> <font face='Lucida Console'>(</font>cluster_id <font color='#5555FF'>=</font><font color='#5555FF'>=</font> labels[j]<font face='Lucida Console'>)</font>
temp.<font color='#BB00BB'>push_back</font><font face='Lucida Console'>(</font>faces[j]<font face='Lucida Console'>)</font>;
<b>}</b>
win_clusters[cluster_id].<font color='#BB00BB'>set_title</font><font face='Lucida Console'>(</font>"<font color='#CC0000'>face cluster </font>" <font color='#5555FF'>+</font> <font color='#BB00BB'>cast_to_string</font><font face='Lucida Console'>(</font>cluster_id<font face='Lucida Console'>)</font><font face='Lucida Console'>)</font>;
win_clusters[cluster_id].<font color='#BB00BB'>set_image</font><font face='Lucida Console'>(</font><font color='#BB00BB'>tile_images</font><font face='Lucida Console'>(</font>temp<font face='Lucida Console'>)</font><font face='Lucida Console'>)</font>;
<b>}</b>
<font color='#009900'>// Finally, let's print one of the face descriptors to the screen.
</font> cout <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> "<font color='#CC0000'>face descriptor for one face: </font>" <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> <font color='#BB00BB'>trans</font><font face='Lucida Console'>(</font>face_descriptors[<font color='#979000'>0</font>]<font face='Lucida Console'>)</font> <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> endl;
<font color='#009900'>// It should also be noted that face recognition accuracy can be improved if jittering
</font> <font color='#009900'>// is used when creating face descriptors. In particular, to get 99.38% on the LFW
</font> <font color='#009900'>// benchmark you need to use the jitter_image() routine to compute the descriptors,
</font> <font color='#009900'>// like so:
</font> matrix<font color='#5555FF'>&lt;</font><font color='#0000FF'><u>float</u></font>,<font color='#979000'>0</font>,<font color='#979000'>1</font><font color='#5555FF'>&gt;</font> face_descriptor <font color='#5555FF'>=</font> <font color='#BB00BB'>mean</font><font face='Lucida Console'>(</font><font color='#BB00BB'>mat</font><font face='Lucida Console'>(</font><font color='#BB00BB'>net</font><font face='Lucida Console'>(</font><font color='#BB00BB'>jitter_image</font><font face='Lucida Console'>(</font>faces[<font color='#979000'>0</font>]<font face='Lucida Console'>)</font><font face='Lucida Console'>)</font><font face='Lucida Console'>)</font><font face='Lucida Console'>)</font>;
cout <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> "<font color='#CC0000'>jittered face descriptor for one face: </font>" <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> <font color='#BB00BB'>trans</font><font face='Lucida Console'>(</font>face_descriptor<font face='Lucida Console'>)</font> <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> endl;
<font color='#009900'>// If you use the model without jittering, as we did when clustering the bald guys, it
</font> <font color='#009900'>// gets an accuracy of 99.13% on the LFW benchmark. So jittering makes the whole
</font> <font color='#009900'>// procedure a little more accurate but makes face descriptor calculation slower.
</font>
cout <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> "<font color='#CC0000'>hit enter to terminate</font>" <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> endl;
cin.<font color='#BB00BB'>get</font><font face='Lucida Console'>(</font><font face='Lucida Console'>)</font>;
<b>}</b>
<font color='#0000FF'>catch</font> <font face='Lucida Console'>(</font>std::exception<font color='#5555FF'>&amp;</font> e<font face='Lucida Console'>)</font>
<b>{</b>
cout <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> e.<font color='#BB00BB'>what</font><font face='Lucida Console'>(</font><font face='Lucida Console'>)</font> <font color='#5555FF'>&lt;</font><font color='#5555FF'>&lt;</font> endl;
<b>}</b>
<font color='#009900'>// ----------------------------------------------------------------------------------------
</font>
std::vector<font color='#5555FF'>&lt;</font>matrix<font color='#5555FF'>&lt;</font>rgb_pixel<font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font> <b><a name='jitter_image'></a>jitter_image</b><font face='Lucida Console'>(</font>
<font color='#0000FF'>const</font> matrix<font color='#5555FF'>&lt;</font>rgb_pixel<font color='#5555FF'>&gt;</font><font color='#5555FF'>&amp;</font> img
<font face='Lucida Console'>)</font>
<b>{</b>
<font color='#009900'>// All this function does is make 100 copies of img, all slightly jittered by being
</font> <font color='#009900'>// zoomed, rotated, and translated a little bit differently. They are also randomly
</font> <font color='#009900'>// mirrored left to right.
</font> thread_local dlib::rand rnd;
std::vector<font color='#5555FF'>&lt;</font>matrix<font color='#5555FF'>&lt;</font>rgb_pixel<font color='#5555FF'>&gt;</font><font color='#5555FF'>&gt;</font> crops;
<font color='#0000FF'>for</font> <font face='Lucida Console'>(</font><font color='#0000FF'><u>int</u></font> i <font color='#5555FF'>=</font> <font color='#979000'>0</font>; i <font color='#5555FF'>&lt;</font> <font color='#979000'>100</font>; <font color='#5555FF'>+</font><font color='#5555FF'>+</font>i<font face='Lucida Console'>)</font>
crops.<font color='#BB00BB'>push_back</font><font face='Lucida Console'>(</font><font color='#BB00BB'>jitter_image</font><font face='Lucida Console'>(</font>img,rnd<font face='Lucida Console'>)</font><font face='Lucida Console'>)</font>;
<font color='#0000FF'>return</font> crops;
<b>}</b>
<font color='#009900'>// ----------------------------------------------------------------------------------------
</font>
</pre></body></html>