News classification
Contact us
- Add: No. 9, North Fourth Ring Road, Haidian District, Beijing. It mainly includes face recognition, living detection, ID card recognition, bank card recognition, business card recognition, license plate recognition, OCR recognition, and intelligent recognition technology.
- Tel: 13146317170 廖经理
- Fax:
- Email: 398017534@qq.com
Face recognition based on depth learning
Face recognition based on depth learning
Face recognition based on depth learning
Preface
Face recognition on the LWF dataset has more than 99.7% face recognition rate. This recognition rate is very high, but how much accuracy is it in the real environment? I don't have any data on this, but I can be sure that the recognition rate in the real environment is not so optimistic. Although there are some commercial applications such as employee recognition management system, customs verification system, and even banks face recognition function, but we can think carefully about the employees face recognition management, so important validation scenarios customs ID card system for identity is actually no businesses boast, for example employees at work brush face if how would fail, is not re identified, or if the wrong recognition or not recognize, is not simply credit card or other registration work, and then scolded his mother, I didn't recognize so handsome! What about the face recognition system on the bank teller machine? Do you dare to let you not even lose your password or brush your face directly? Do you turn off the face recognition and fingerprint recognition machine and run it properly? So in the light of various factors, age factors, reticulocyte factors in real environment (makeup) and even identify cheating factors and other factors under the condition of the rate of how many manufacturers only know their own, I believe that each manufacturer for these situations are optimized, such as auxiliary peripheral hardware, for a particular scene all kinds of constraints and so on, through the various manufacturers to optimize the system in all aspects, can enhance the experience of their products integrated.
The front is far away. The purpose of this paper is to achieve the simplest practical application of face recognition, that is, to capture dynamic faces with cameras, and then compare the 128D face features that have been stored in the database to identify the corresponding face information (names, etc.).
In this system, I have stored the face features of the front head images of several stars in advance, and of course you can store and import more faces, of course.
Then through the face detection, face image processing, face recognition and other steps to identify the corresponding face information, the recognition effect is as follows
Of course, this is just a simple application, the real use of production system needed by in vivo detection, avoid the use of photos or video, mobile phone cheated face recognition system, application of higher level of security requirements such as payment, transfer system for in vivo detection may still not safe enough, it can also strengthen safety the performance of face recognition + password etc..
Face database import
Face data import, that is to say I start at the beginning of the system, need to import my face database, that is the front of those stars of the front. In the initial stage of loading, because we need to detect the face part of static face images, we first need to use Dlib's face detector and get_frontal_face_detector () to get it. Then you need to import the 68 point face markup model into shape_predictor SP, in order to pose to the person's face in a standard position and then to load the DNN model. Then we take every feature of facial photos and put information related to features and names into FACE_DESC structure, and finally put each face information structure in face_desc_vec container. Here I only load 9 star's face information.
[cpp] view plain copy
Int FACE_RECOGNITION:: load_db_faces (void)
{
INTRC = -1;
LonghFile = 0;
Struct_finddata_tfileinfo;
Frontal_face_detectordetector =get_frontal_face_detector ();
We will also use a face / landmarking model to align faces to a standard pose: (see face_landmark_detection_ex.cpp for an introduction)
Deserialize ("shape_predictor_68_face_landmarks.dat") >>sp;
And finally we load the DNN / responsible for face recognition.
Deserialize ("dlib_face_recognition_resnet_model_v1.dat") >>net;
If ((hFile =_findfirst (".\\faces\\*.jpg", &fileinfo))! = -1)
{
Do
{
If ((fileinfo.attrib &_A_ARCH))
{
If (StrCmp ("fileinfo.name")! = 0 & & StrCmp (fileinfo.name, "..")! = 0)
{
If (! StrCmp (strstr (fileinfo.name, ".") + 1, "JPG"))
{
"Cout" "This file is an image file!" <
Matriximg;
Charpath[260];
Sprintf_s (path, ".\\faces\\%s", fileinfo.name);
Load_image (IMG, path);
Image_windowwin (IMG);
For (autoface: detector (IMG))
{
AutoShape =sp (IMG, face);
Matrixface_chip;
Extract_image_chip (IMG, get_face_chip_details (shape, 150, 0.25), face_chip);
//Record the all this face's information
FACE_DESCsigle_face;
Sigle_face.face_chip =face_chip;
Sigle_face.name =fileinfo.name;
Std:: vectorface_chip_vec;
Std:: vector
Preface
Face recognition on the LWF dataset has more than 99.7% face recognition rate. This recognition rate is very high, but how much accuracy is it in the real environment? I don't have any data on this, but I can be sure that the recognition rate in the real environment is not so optimistic. Although there are some commercial applications such as employee recognition management system, customs verification system, and even banks face recognition function, but we can think carefully about the employees face recognition management, so important validation scenarios customs ID card system for identity is actually no businesses boast, for example employees at work brush face if how would fail, is not re identified, or if the wrong recognition or not recognize, is not simply credit card or other registration work, and then scolded his mother, I didn't recognize so handsome! What about the face recognition system on the bank teller machine? Do you dare to let you not even lose your password or brush your face directly? Do you turn off the face recognition and fingerprint recognition machine and run it properly? So in the light of various factors, age factors, reticulocyte factors in real environment (makeup) and even identify cheating factors and other factors under the condition of the rate of how many manufacturers only know their own, I believe that each manufacturer for these situations are optimized, such as auxiliary peripheral hardware, for a particular scene all kinds of constraints and so on, through the various manufacturers to optimize the system in all aspects, can enhance the experience of their products integrated.
The front is far away. The purpose of this paper is to achieve the simplest practical application of face recognition, that is, to capture dynamic faces with cameras, and then compare the 128D face features that have been stored in the database to identify the corresponding face information (names, etc.).
In this system, I have stored the face features of the front head images of several stars in advance, and of course you can store and import more faces, of course.
Then through the face detection, face image processing, face recognition and other steps to identify the corresponding face information, the recognition effect is as follows
Of course, this is just a simple application, the real use of production system needed by in vivo detection, avoid the use of photos or video, mobile phone cheated face recognition system, application of higher level of security requirements such as payment, transfer system for in vivo detection may still not safe enough, it can also strengthen safety the performance of face recognition + password etc..
Face database import
Face data import, that is to say I start at the beginning of the system, need to import my face database, that is the front of those stars of the front. In the initial stage of loading, because we need to detect the face part of static face images, we first need to use Dlib's face detector and get_frontal_face_detector () to get it. Then you need to import the 68 point face markup model into shape_predictor SP, in order to pose to the person's face in a standard position and then to load the DNN model. Then we take every feature of facial photos and put information related to features and names into FACE_DESC structure, and finally put each face information structure in face_desc_vec container. Here I only load 9 star's face information.
[cpp] view plain copy
Int FACE_RECOGNITION:: load_db_faces (void)
{
INTRC = -1;
LonghFile = 0;
Struct_finddata_tfileinfo;
Frontal_face_detectordetector =get_frontal_face_detector ();
We will also use a face / landmarking model to align faces to a standard pose: (see face_landmark_detection_ex.cpp for an introduction)
Deserialize ("shape_predictor_68_face_landmarks.dat") >>sp;
And finally we load the DNN / responsible for face recognition.
Deserialize ("dlib_face_recognition_resnet_model_v1.dat") >>net;
If ((hFile =_findfirst (".\\faces\\*.jpg", &fileinfo))! = -1)
{
Do
{
If ((fileinfo.attrib &_A_ARCH))
{
If (StrCmp ("fileinfo.name")! = 0 & & StrCmp (fileinfo.name, "..")! = 0)
{
If (! StrCmp (strstr (fileinfo.name, ".") + 1, "JPG"))
{
"Cout" "This file is an image file!" <
Matriximg;
Charpath[260];
Sprintf_s (path, ".\\faces\\%s", fileinfo.name);
Load_image (IMG, path);
Image_windowwin (IMG);
For (autoface: detector (IMG))
{
AutoShape =sp (IMG, face);
Matrixface_chip;
Extract_image_chip (IMG, get_face_chip_details (shape, 150, 0.25), face_chip);
//Record the all this face's information
FACE_DESCsigle_face;
Sigle_face.face_chip =face_chip;
Sigle_face.name =fileinfo.name;
Std:: vectorface_chip_vec;
Std:: vector