Autolocker

Facial recognition-based PC access control tool.

This tool uses a camera to detect the face of a person using the PC and validate it against a pre-generated set of faces of authorized users. If the face of a person using the PC is not present in the set or if there's nobody in front of the camera, then PC is 'locked down' with full-screen window which dissapears once an authorized face is recognized.

Usability

The tool isn't perfectly suitable for continuous daily use because Local Binary Patterns Historgrams algorithm turned out to be quite sensitive to the face alignment, lighting conditions and image quality, especially with small sets of reference images. As a result, in order to avoid annoying short lockdowns due to failed recognitions, user has to keep his head somewhat straight at all times and that seems to build a lot of uncomfortable tension in the user's neck.

Requirements

OpenCV 3.4.1 built with contrib libraries
Visual Studio 15 2017 with C++ development environment

How to build OpenCV

Before you start: it's still not too late to turn back. Don't be surprised when you find yourself cursing at your PC tommorrow's early morning. You've been warned.

Now, with legal shenaningans and common courtesies out of the way, let's start with OpenCV's own tutorial. Take a look at the Installation using Git-bash and cmake section. PROTIP: you might want to provision and dedicate VM and slap VS 2017 Community into it before proceeding with building, things are going to get messy.

Here's the copy of OpenCV's tutorial in case if it's no longer there.

Download and install cmake (3.9.1 or higher). Make sure to add cmake to PATH during the installation.
Install git-bash (2.14.1 or higher). Do not add git to PATH variable during the installation.
Run git-bash, you should see a command window.
In git-bash command window enter following commands (mind the path, at this point you're basically selecting where do you want OpenCV to be installed):

mkdir /c/lib
cd /c/lib

Copy the script below and save it into C:\lib under name installOCV.sh:

#!/bin/bash -e
myRepo=$(pwd)
CMAKE_CONFIG_GENERATOR="Visual Studio 15 2017 Win64"
if [  ! -d "$myRepo/opencv"  ]; then
    echo "clonning opencv"
    git clone https://github.com/opencv/opencv.git
    mkdir Build
    mkdir Build/opencv
    mkdir Install
    mkdir Install/opencv
else
    cd opencv
    git pull --rebase
    cd ..
fi
if [  ! -d "$myRepo/opencv_contrib"  ]; then
    echo "clonning opencv_contrib"
    git clone https://github.com/opencv/opencv_contrib.git
    # mkdir Build < TRY UNCOMMENTING THAT LINE IF THINGS ARE FAILING
    mkdir Build/opencv_contrib
else
    cd opencv_contrib
    git pull --rebase
    cd ..
fi
RepoSource=opencv
pushd Build/$RepoSource
CMAKE_OPTIONS='-DBUILD_PERF_TESTS:BOOL=OFF -DBUILD_TESTS:BOOL=OFF -DBUILD_DOCS:BOOL=OFF  -DWITH_CUDA:BOOL=OFF -DBUILD_EXAMPLES:BOOL=OFF -DINSTALL_CREATE_DISTRIB=ON'
cmake -G"$CMAKE_CONFIG_GENERATOR" $CMAKE_OPTIONS -DOPENCV_EXTRA_MODULES_PATH="$myRepo"/opencv_contrib/modules -DCMAKE_INSTALL_PREFIX="$myRepo"/install/"$RepoSource" "$myRepo/$RepoSource"
echo "************************* $Source_DIR -->debug"
cmake --build .  --config debug
echo "************************* $Source_DIR -->release"
cmake --build .  --config release
cmake --build .  --target install --config release
cmake --build .  --target install --config debug
popd

This script supposes that you use Visual Studio 2017 in 64 bits

CMAKE_CONFIG_GENERATOR="Visual Studio 15 2017 Win64"

and that the OpenCV will be envetually built into C:\lib\Install:

-DCMAKE_INSTALL_PREFIX="$myRepo"/install/"$RepoSource" "$myRepo/$RepoSource"

with no performance tests, unit test, no documentation, no CUDA support and no examples:

CMAKE_OPTIONS='-DBUILD_PERF_TESTS:BOOL=OFF -DBUILD_TESTS:BOOL=OFF -DBUILD_DOCS:BOOL=OFF -DWITH_CUDA:BOOL=OFF -DBUILD_EXAMPLES:BOOL=OFF'

In git-bash command line enter the following command:

./installOCV.sh

Try to solve the errors. God help you.
Eventually you'll have the thing building, it will take a while.
(Not sure if required) Once Cmake finishes building the project, navigate into C:\lib\Build, open the .sln file and build the entire solution.

How to install OpenCV

This section assumes that you've built the OpenCV with contrib stuff by following the section above.

Long story short, set the OpenCV enironment variable and add it to the system's path.

Long story long,

Open Environment variables window (see System Properties)
Under System variables group, create an OPENCV_DIR variable and set it to the location of your OpenCV's bin and lib folders, something like that: C:\Program Files (x86)\OpenCV\opencv\build\x64\vc15.
Under System variables group, find an existing PATH variable and click 'Edit'. In appeared dialog, add a new path with exactly the following value: %OPENCV_DIR%\bin.
Reboot your PC. After that you should be all set.

How to build the tool

Simply build the solution as you normally would using Visual Studio.

Targeting the newer version of OpenCV

Update the following project properties:

Additional Dependencies (under Linker > Input)
Additional Library Directories (under Linker > General)
Additional Include Directories (under C/C++ > General)
Path to opencv_world<version>.dll (under Build Events > Post-Build Event > Command Line)

Required stuff to launch an executable

AutoLocker.exe (duh)
opencv_world341.dll
'patterns' folder with 'haarcascade_frontalface_alt.xml' file
'authorized_face' folder with 'reference_face.jpg' file

All these files will be automatically copied into Package directory within the build destination during post-build step.

Add an authorized user

Run the Autolocker.exe and type learn.
Wait until the program captures and shows you an image from camera.
If an image is vaild, type Y to save it or N to repeat (this prompt might be hidden behind image window).

The process will be repeated five times. Feel free to run learn when you feel that existing photos aren't allowing the program to recognize the user (due to different lighting conditions or whatnot). However, be mindful of amount of images the application has to load as it can easily chew through large amounts of RAM with relatively low image count (2GB of RAM for 90 images).

Remove an authorized user

Navigate to ./authorized_faces folder and remove images manually.

Settings description

Set settings by typing sm set <parameter name> <value>

engine_rpm - a delay in milliseconds between engine iterations.
detectionfailure_threshold - amount of consequental failed face detections after which lockdown is engaged.
recognitionfailure_threshold - amount of consequental failed face recognitions after which lockdown is engaged.
recognition_interval - amount of seconds between each recognition.
confidence_threshold - a recognition confidence value after which a face is considered unrecognized. Lower values mean higher recognition confidence.
capturedevide_index - an index of a capture device. Usually 0.
logverbosity - log verbosity level. Acceptable values (following levels include preceeding ones): 0 - Silent, 1 - Warnings, 2 - Errors, 3 - Logs, 4 - Debug.
feedwindow - boolean switch that specifies if camera feed must be show during program's operation.
preventlockdown - boolean switch that allows to prevent actual lockdown. Useful for thershold calibration.
cascadetemplate_path - a path to the template consumed by face detector.
authorizedfaces_path - a path tho the folder with images of authorized users.

igprya/Autolocker