All Projects → IntelRealSense → RealSenseID

IntelRealSense / RealSenseID

Licence: Apache-2.0 license
Intel® RealSense™ ID SDK

Programming Languages

C++
36643 projects - #6 most used programming language
C#
18002 projects
java
68154 projects - #9 most used programming language
c
50402 projects - #5 most used programming language
CMake
9771 projects
powershell
5483 projects
shell
77523 projects

Projects that are alternatives of or similar to RealSenseID

FaceNet-based-Attendance-System
Deep Learning based Web Application for marking attendance of students by recognizing the student's faces from the surveillance video footage of classroom.
Stars: ✭ 25 (-62.69%)
Mutual labels:  face-recognition
formulas-python
Ritchie CLI formulas in Python 🐍
Stars: ✭ 17 (-74.63%)
Mutual labels:  face-recognition
shunyaface
Fast Face Recognition on the edge
Stars: ✭ 49 (-26.87%)
Mutual labels:  face-recognition
lbph
Local Binary Patterns Histograms (LBPH) implementation in Go
Stars: ✭ 50 (-25.37%)
Mutual labels:  face-recognition
Look4Face
Demo of Face Recognition web service
Stars: ✭ 23 (-65.67%)
Mutual labels:  face-recognition
DeepLearning-Gdansk2019-tutorial
Ordinal Regression tutorial for the International Summer School on Deep Learning 2019
Stars: ✭ 66 (-1.49%)
Mutual labels:  face-recognition
PassFace
Face recognizing program develop using EmguCV (OpenCV wrapper). It is possible to use 4 different algorithm and 3 different source(Camera, Video, Image)
Stars: ✭ 38 (-43.28%)
Mutual labels:  face-recognition
ElasticFace
Official repository for ElasticFace: Elastic Margin Loss for Deep Face Recognition
Stars: ✭ 86 (+28.36%)
Mutual labels:  face-recognition
AIML-Human-Attributes-Detection-with-Facial-Feature-Extraction
This is a Human Attributes Detection program with facial features extraction. It detects facial coordinates using FaceNet model and uses MXNet facial attribute extraction model for extracting 40 types of facial attributes. This solution also detects Emotion, Age and Gender along with facial attributes.
Stars: ✭ 48 (-28.36%)
Mutual labels:  face-recognition
Face-X
Demonstration of different algorithms and operations on faces. Join the Discord channel for discussion.
Stars: ✭ 226 (+237.31%)
Mutual labels:  face-recognition
age-and-gender
Predict Age and Gender of people from images | Determination of gender and age
Stars: ✭ 68 (+1.49%)
Mutual labels:  face-recognition
WhoAreYou
Face detection and recognition with CoreML and ARKit
Stars: ✭ 91 (+35.82%)
Mutual labels:  face-recognition
face-swap
换脸程序
Stars: ✭ 32 (-52.24%)
Mutual labels:  face-recognition
WreckFaceApp
Android application for gender, age and face recognition using OpenCV and JavaCV libraries
Stars: ✭ 21 (-68.66%)
Mutual labels:  face-recognition
esp32cam-access-control
Open a door when a face is recognised using the ESP32-CAM
Stars: ✭ 45 (-32.84%)
Mutual labels:  face-recognition
face-recognition
🙂 An OpenCV application that can detect & recognize your face in real-time
Stars: ✭ 13 (-80.6%)
Mutual labels:  face-recognition
FaceClassification Tensorflow
Building a Neural Network that classifies faces using OpenCV and Tensorflow
Stars: ✭ 37 (-44.78%)
Mutual labels:  face-recognition
pigallery
PiGallery: AI-powered Self-hosted Secure Multi-user Image Gallery and Detailed Image analysis using Machine Learning, EXIF Parsing and Geo Tagging
Stars: ✭ 35 (-47.76%)
Mutual labels:  face-recognition
CS-663
Assignment Codes for CS663 Digital Image Processing
Stars: ✭ 15 (-77.61%)
Mutual labels:  face-recognition
facerec-bias-bfw
Source code and notebooks to reproduce experiments and benchmarks on Bias Faces in the Wild (BFW).
Stars: ✭ 40 (-40.3%)
Mutual labels:  face-recognition

Intel® RealSense™ ID Solution for Facial Authentication

Overview

Intel RealSense ID is your trusted facial authentication on-device solution.

Intel RealSense ID combines an active depth sensor with a specialized neural network designed to deliver an intuitive, secure and accurate facial authentication solution that adapts over time. This solution offers user privacy and is activated by user awareness. Built-in anti spoofing technology protects against attacks using photographs, videos or masks.

Intel RealSense ID is a natural solution simplifying secure entry for everyone, everywhere. Supports children to tall adults and designed for Smart Locks, Access Control, PoS, ATMs, and Kiosks.

Developer kits containing the necessary hardware to use this library are available for purchase at store.intelrealsense.com. See information about the Intel RealSense ID technology at https://www.intelrealsense.com/facial-authentication/

For high-level architecture, see Intel RealSense ID F450 / F455 Architecture Diagram.

Note: Device = Intel RealSense ID F450 / F455

Platforms

  • Linux (tested on ubuntu 18, gcc 7.5+)
  • Windows (tested on windows 10, msvc 2019)
  • Android (tested on Android 6.0 but should also work on newer versions)

Building

Use CMake version 3.10.2 or above:

$ cd <project_dir>
$ mkdir build && cd build
$ cmake .. //in case preview is required run: cmake .. -DRSID_PREVIEW=1 //in case secure is required run: cmake .. -DRSID_SECURE=1
$ make -j

Sample Code

This snippet shows the basic usage of our library.

// Create face authenticator and connect it to the device on port COM9
RealSenseID::FaceAuthenticator authenticator;
connect_status = authenticator.Connect({RealSenseID::SerialType::USB, "COM9"});

// Enroll a user
const char* user_id = "John";
MyEnrollClbk enroll_clbk;
auto status = authenticator.Enroll(enroll_clbk, user_id);

// Authenticate a user
MyAuthClbk auth_clbk;
status = authenticator.Authenticate(auth_clbk);

// Remove the user from device
success = authenticator.RemoveUser(user_id);

For additional languages, build instruction and detailed code please see our code samples and tools.

Secure Communication

The library can be compiled in secure mode. Once paired with the device, all communications will be protected. If you wish to get back to non-secure communcations, you must first unpair your device.

Cryptographic algorithms that are used for session protection:

  • Message signing - ECDSA P-256 (secp256-r1).
  • Shared session key generation - ECDH P-256 (secp256-r1)
  • Message encryption - AES-256, CTR mode.
  • Message integrity - HKDF-256.

To enable secure mode, the host should perform the following steps:

  • Compile the library with RSID_SECURE enabled.
  • Generate a set of ECDSA P-256 (secp256-r1) public and private keys. The host is responsible for keeping his private key safe.
  • Pair with the device to enable secure communication. Pairing is performed once using the FaceAuthenticator API.
  • Implement a SignatureCallback. Signing and verifying is done with the ECDSA P-256 (secp256-r1) keys. Please see the pairing sample on how to pair the device and use keys.

Each request (a call to one of the main API functions listed below) from the host to the device starts a new encrypted session, which performs an ECDH P-256 (secp256-r1) key exchange to create the shared secret for encrypting the current session.

class MySigClbk : public RealSenseID::SignatureCallback
{
public:
    bool Sign(const unsigned char* buffer, const unsigned int buffer_len, unsigned char* out_sig) override
    {
      // Sign buffer with host ECDSA private key
    }

    bool Verify(const unsigned char* buffer, const unsigned int buffer_len, const unsigned char* sig,const unsigned int sign_len) override
    {
      // Verify buffer with device ECDSA public key
    }
};

MySigClbk sig_clbk;
RealSenseID::FaceAuthenticator authenticator {&sig_clbk};
Status connect_status = authenticator.Connect({RealSenseID::SerialType::USB, "COM9"});
authenticator.Disconnect();
char* hostPubKey = getHostPubKey(); // 64 bytes ECDSA public key of the host
char host_pubkey_signature[64];
if (!sig_clbk.Sign((unsigned char*)hostPubKey, 64, host_pubkey_signature))
{
    printf("Failed to sign host ECDSA key\n");
    return;
}
char device_pubkey[64]; // Save this key after getting it to use it for verification
Status pair_status = authenticator.Pair(host_pubkey, host_pubkey_signature, device_pubkey);

Server and Device APIs

Device Mode

Device Mode is a set of APIs enable the user to enroll and authenticate on the device itself, including database management and matching on the device.

Server Mode

Server Mode is a set of APIs for users who wish to manage a faceprints database on the host or the server. In this mode F450 is used as a Faceprints-Extraction device only.

The API provides a 'matching' function which predicts whether two faceprints belong to the same person, thus enabling the user to scan their database for similar users.

Project Structure

RealSenseID API

The RealSenseID is the main API to the communication with the device. C, C++ and C# API wrappers are provided as part of the library. Java and python API is under development. All APIs are synchronous and assuming a new operation will not be activated till previous is completed.

FaceAuthenticator

Connect / Disconnect

Connects host to device using USB (RealSenseID::SerialType::USB) or UART interfaces (RealSenseID::SerialType::UART).

class MySigClbk : public RealSenseID::SignatureCallback
{
public:
    bool Sign(const unsigned char* buffer, const unsigned int buffer_len, unsigned char* out_sig) override
    {
      // Sign buffer with host ECDSA private key
    }

    bool Verify(const unsigned char* buffer, const unsigned int buffer_len, const unsigned char* sig,const unsigned int sign_len) override
    {
      // Verify buffer with device ECDSA public key
    }
};

MySigClbk sig_clbk;
RealSenseID::FaceAuthenticator authenticator {&sig_clbk};
Status connect_status = authenticator.Connect({RealSenseID::SerialType::USB, "COM9"});
authenticator.Disconnect();
Pair

Perform an ECDSA key exchange. In pairing, the host sends the following:

  • ECDSA public key.
  • ECDSA public key signature with the previous ECDSA private key. If this is the first pairing, the signature is ignored.
Status pair_status = authenticator.Pair(host_pubkey, host_pubkey_signature, device_pubkey);

The Host will receive the device's ECDSA public key, and should save it for verification in the implementation of SignatureCallback::Verify.

Enrollment

Starts device, runs neural network algorithm and stores encrypted faceprints on database. A faceprint is a set number of points which is represented as mathematical transformation of the user’s face. saves encrypted facial features to secured flash on Intel RealSense ID F450 / F455. Stored encrypted faceprints are matched with enrolled faceprints later during authentication. For best performance, enroll under normal lighting conditions and look directly at the device. During the enrollment process, device will send a status hint to the callback provided by application. Full list of the hint can be found in EnrollStatus.h.

class MyEnrollClbk : public RealSenseID::EnrollmentCallback
{
public:
    void OnResult(const RealSenseID::EnrollStatus status) override
    {
        std::cout << "on_result: status: " << status << std::endl;
    }

    void OnProgress(const RealSenseID::FacePose pose) override
    {
        std::cout << "on_progress: pose: " << pose << std::endl;
    }

    void OnHint(const RealSenseID::EnrollStatus hint) override
    {
        std::cout << "on_hint: hint: " << hint << std::endl;
    }
};

const char* user_id = "John";
MyEnrollClbk enroll_clbk;
Status status = authenticator.Enroll(enroll_clbk, user_id);
Authenticate

Single authentication attempt: Starts device, runs neural network algorithm, generates faceprints and compares them to all enrolled faceprints in database. Finally, returns whether the authentication was forbidden or allowed with enrolled user id. During the authentication process, device will send a status hint to the callback provided by application. Full list of the hint can be found in AuthenticationStatus.h.

This operation can be further configured by passing DeviceConfig struct to the FaceAuthenticator::SetDeviceConfig(const DeviceConfig&) function. See the Device Configuration below for details.

class MyAuthClbk : public RealSenseID::AuthenticationCallback
{
public:
    // Called when authentication result is available.
    // If there are multiple faces it will be called for each face detected.
    void OnResult(const RealSenseID::AuthenticateStatus status, const char* user_id) override
    {
        if (status == RealSenseID::AuthenticateStatus::Success)
        {
            std::cout << "******* Authenticate success.  user_id: " << user_id << " *******" << std::endl;
        }
        else
        {
            std::cout << "on_result: status: " << status << std::endl;
        }
    }

    void OnHint(const RealSenseID::AuthenticateStatus hint) override
    {
        std::cout << "on_hint: hint: " << hint << std::endl;
    }
};

MyAuthClbk auth_clbk;
Status status = authenticator.Authenticate(auth_clbk);
AuthenticateLoop

Starts device, runs authentication in a loop until Cancel() API is invoked. Starts camera and running loop of

  • NN algo(s) pipeline
  • Facial features extraction
  • Matches them to all previously stored features in database.
  • Returns result whether the authentication was forbidden or allowed with pre-enrolled user id. During the authentication process hint statuses might be sent to the callback provided by application. Full list of the hints can be found in AuthenticationStatus.h.
class MyAuthClbk : public RealSenseID::AuthenticationCallback
{
public:
    // Called when result is available for a detected face.
    // If the status==AuthenticateStatus::Success then the user_id will point to c string of the authenenticated user id.
    void OnResult(const RealSenseID::AuthenticateStatus status, const char* user_id) override
    {
        if (status == RealSenseID::AuthenticateStatus::Success)
        {
            std::cout << "******* Authenticate success.  user_id: " << user_id << " *******" << std::endl;
        }
        else
        {
            std::cout << "on_result: status: " << status << std::endl;
        }
    }

    void OnHint(const RealSenseID::AuthenticateStatus hint) override
    {
        std::cout << "on_hint: hint: " << hint << std::endl;
    }

   
    // Called when faces are detected. Can be single or multiple faces.
    // The faces vector contains coord X(,y,h,w)of faces detected 
    // Coords are in full HD resolution (1920x1080):
    //   x,y: top left face rect
    //   w,h: width/height of the face rect
    //   Note: The `X` coords are flipped which means X==0 is most right, and X==1920 is most left.         
    // The timestamp argument is the timestamp in millis of the frame that the faces were found in.    
    void OnFaceDetected(const std::vector<RealSenseID::FaceRect>& faces, const unsigned int timestamp) override
    {
        _faces = faces;
    }
};

MyAuthClbk auth_clbk;
Status status = authenticator.AuthenticateLoop(auth_clbk);
Device Configuration API

The device operation can be configured by passing the DeviceConfig struct to the FaceAuthenticator::SetDeviceConfig(const DeviceConfig&) function.

The various options and default values are described below:

struct RSID_API DeviceConfig
{
    /**
     * @enum CameraRotation
     * @brief Camera rotation.
     */
    enum class CameraRotation
    {
        Rotation_0_Deg = 0, // default
        Rotation_180_Deg = 1,
        Rotation_90_deg = 2,
        Rotation_270_deg = 3
    };

    /**
     * @enum SecurityLevel
     * @brief SecurityLevel to allow
     */
    enum class SecurityLevel
    {
        High = 0,   // high security, no mask support, all AS algo(s) will be activated
        Medium = 1, // default mode to support masks, only main AS algo will be activated.
    };
  
    /**
     * @enum AlgoFlow
     * @brief Algorithms which will be used during authentication
     */
    enum class AlgoFlow
    {
        All = 0,               // default
        FaceDetectionOnly = 1, // face detection only
        SpoofOnly = 2,         // spoof only
        RecognitionOnly = 3    // recognition only
    };

    /**
     * @enum FaceSelectionPolicy
     * @brief To run authentication on all (up to 5) detected faces vs single (closest) face
     */
    enum class FaceSelectionPolicy
    {
        Single = 0, // default, run authentication on closest face
        All = 1     // run authentication on all (up to 5) detected faces
    };

    enum class DumpMode
    {
        None = 0,
        CroppedFace = 1,
        FullFrame = 2,
    };

    CameraRotation camera_rotation = CameraRotation::Rotation_0_Deg;
    SecurityLevel security_level = SecurityLevel::Medium;
    AlgoFlow algo_flow = AlgoFlow::All;
    FaceSelectionPolicy face_selection_policy = FaceSelectionPolicy::Single;
    DumpMode dump_mode = DumpMode::None;
};

Notes:

  • If SetDeviceConfig() never called, the device will use the default values described above.
  • SetDeviceConfig() can be called once. The settings will take effect for all future authentication sessions (until the device is restarted).
  • CameraRotation enable the algorithm to work with a rotated device. For preview rotation to match, you'll need to define previewConfig.portraitMode accordingly (see Preview section).

The following example configures the device to only detect spoofs (instead of the default full authentication):

...
using namespace RealSenseID;
// config with default options except for the algo flow which is set to spoof only 
DeviceConfig device_config; 
device_config.algo_flow = AlgoFlow::SpoofOnly;
auto status = authenticator->SetDeviceConfig(device_config);
Cancel

Stops current operation (Enrollment/Authenticate/AuthenticateLoop).

authenticator.Cancel();
RemoveAll

Removes all enrolled users in the device database.

bool success = authenticator.RemoveAll();
RemoveUser

Removes a specific user from the device database.

const char* user_id = "John";
bool success = authenticator.RemoveUser(user_id);

DeviceController

Connect / Disconnect

Connects host to device using USB (RealSenseID::SerialType::USB) or UART interfaces (RealSenseID::SerialType::UART).

RealSenseID::DeviceController deviceController;
Status connect_status = deviceController.Connect({RealSenseID::SerialType::USB, "COM9"});
deviceController.Disconnect();
Reboot

Resets and reboots device.

bool success = deviceController.Reboot();
FW Upgrade (Not supported yet)

Preview API

Currently 704x1280 or 1056x1920 RGB formats is available.

Preview Configuration
/**
 * Preview modes
 */
enum class PreviewMode
{
    MJPEG_1080P = 0, // default
    MJPEG_720P = 1,
    RAW10_1080P = 2,
};

/**
 * Preview configuration
 */
struct RSID_API PreviewConfig
{
    int cameraNumber = -1; // attempt to auto detect by default
    PreviewMode previewMode = PreviewMode::MJPEG_1080P; // RAW10 requires custom fw support
    bool portraitMode = true; 
};

Notes:

  • The rotation used by algorithm is based only on DeviceConfig.camera_rotation attribute.
  • Indepedently, you can choose each preview mode (except raw) to be portrait or non-portrait.
  • Keep in mind that if you want preview to match algo:
    CameraRotation::Rotation_0_Deg and CameraRotation::Rotation_180_Deg is for portraitMode == true.(default)
    CameraRotation::Rotation_90_Deg and CameraRotation::Rotation_270_Deg is for portraitMode == false.
Sensor Timestamps

Access to the sensor timestamps (in milliseconds).
To enable it on Windows please turn on the Metadata option in when using the SDK installer. The installer create specific dedicated registry entry to be present for each unique RealSenseID device. For Linux, Metadata is supported on kernels 4.16 + only.

The timestamps can be acquired in OnPreviewImageReady under image.metadata.timestamp . Other metadata isn't valid.

more information about metadata on Windows can be found in microsoft uvc documetnation

StartPreview

Starts preview. Callback function that is provided as parameter will be invoked for a newly arrived image and can be rendered by your application.

class PreviewRender : public RealSenseID::PreviewImageReadyCallback
{
public:
	void OnPreviewImageReady(const Image image)
	{
		// Callback will be used to provide RGB preview image (for RAW10_1080P PreviewMode - raw converted to RGB).
	}

    	void OnSnapshotImageReady(const Image image) // Not mandatory. To enable it, see Device Configuration API.
	{
		// Callback will be used to provide images destined to be dumped (cropped/full). See DeviceConfig::DumpMode attribute.
		// Raw formatted images will be raised by this function.
	}
};

PreviewRender image_clbk;
Preview preview;
bool success = preview.StartPreview(image_clbk);
PausePreview

Pause preview.

bool success = preview.PausePreview();
ResumePreview

Resumes the preview.

bool success = preview.ResumePreview();
StopPreview

Stops preview.

bool success = preview.StopPreview();

Server Mode Methods

ExtractFaceprintsForAuth

Extracts faceprints from a face, in the device, and sends them to the host. Uses 'Authentication Flow' to eliminate spoof attempts and verify a face was detected.

// extract faceprints using authentication flow
class FaceprintsAuthClbk : public RealSenseID::AuthFaceprintsExtractionCallback
{
public:
    void OnResult(const RealSenseID::AuthenticateStatus status, const Faceprints* faceprints) override
    {
        std::cout << "result: " << status << std::endl;
        s_extraction_status = status;
        // if status was success pass '_faceprints' to user 
    }

    void OnHint(const RealSenseID::AuthenticateStatus hint) override
    {
        std::cout << "hint: " << hint << std::endl;
    }
};

FaceprintsAuthClbk clbk;
auto status = authenticator->ExtractFaceprintsForAuth(clbk);
ExtractFaceprintsForEnroll

Extracts faceprints from a face, in the device, and sends them to the host. Uses 'Enrollment Flow' to verify face-pose is correct and that a face was detected.

// extract faceprints using enrollment flow
class MyEnrollServerClbk : public RealSenseID::EnrollmentCallback
{
    const char* _user_id = nullptr;

public:
    explicit MyEnrollServerClbk(const char* user_id) : _user_id {user_id}
    {        
    }

    void OnResult(const RealSenseID::EnrollStatus status, const Faceprints* faceprints) override
    {
        std::cout << "result: " << status << "for user: " << _user_id << std::endl;
        s_extraction_status = status;    
        // if status was success pass '_faceprints' to user            
    }

    void OnProgress(const RealSenseID::FacePose pose) override
    {
        std::cout << "pose: " << pose << std::endl;
    }

    void OnHint(const RealSenseID::EnrollStatus hint) override
    {
        std::cout << "hint: " << hint << std::endl;
    }
};

MyEnrollServerClbk enroll_clbk {user_id.c_str()};
auto status = authenticator->ExtractFaceprintsForEnroll(enroll_clbk);
ExtractFaceprintsForAuthLoop

Extracts faceprints in a loop, each iteration extracts from a single face and sends it to the host. Uses 'Authentication Flow' to eliminate spoof attempts and verify a face was detected.

// extract faceprints in a loop using authentication flow
class AuthLoopExtrClbk : public RealSenseID::AuthFaceprintsExtractionCallback
{    
public:
    void OnResult(const RealSenseID::AuthenticateStatus status, const Faceprints* faceprints) override
    {        
        std::cout << "result: " << status << std::endl;
        s_extraction_status = status;
        // if status was success pass '_faceprints' to user 
    }

    void OnHint(const RealSenseID::AuthenticateStatus hint) override
    {
        std::cout << "hint: " << hint << std::endl;
    }
};

AuthLoopExtrClbk clbk;
auto status = authenticator->ExtractFaceprintsForAuthLoop(clbk);
MatchFaceprints

Matches two faceprints to each other and calculates a prediction of whether they belong to the same person.

// match extracted faceprints to the ones in the user-managed database
for (auto& db_item : db)
{
    auto match_result = authenticator->MatchFaceprints(scanned_faceprints, db_item.faceprints, updated_faceprints);    
    if (match_result.success)
    {
        std::cout << "Match succeeded with user id: " << db_item.id << std::endl;
        break;
    }
}

Python API (new)

Starting with version 0.22.0 we provide python API:

""" 
Authenticate example
"""
import rsid_py

def on_result(result, user_id):    
    if result == rsid_py.AuthenticateStatus.Success:
        print('Authenticated user:', user_id)

with rsid_py.FaceAuthenticator("COM9") as f:
    f.authenticate(on_result=on_result)

Please visit the python samples page for details and samples.

PacketManager

The PacketManager is the communication protocol implementation. Communication is encrypted and managed by PacketManager::SecureSessionManager.

SecureSessionManager

Manages the secure session. Sends packets using PacketManager::PacketSender.

PacketSender

Responsible for sending and receiving serial packets over a serial port using PacketManager::SerialConnection interface.

SerialConnection

Interface for communication over a serial port. Implemented over supported platforms (WindowsSerial/LinuxSerial/Android). For new platform, implement this interface.

SerialPacket

Packet structure base. We have 2 types of packets:

  • DataPacket - Payload contains data buffer.
  • FaPacket - Payload contains user id and status.

License

This project is licensed under Apache 2.0 license. Relevant license info can be found in "License Notices" folder.

Intel RealSense ID F450 and F455 Architecure Diagram

plot

Note that the project description data, including the texts, logos, images, and/or trademarks, for each open source project belongs to its rightful owner. If you wish to add or remove any projects, please contact us at [email protected].