/biometric_camera_signin

Camera-based facial biometric sign in application

Primary LanguagePythonBSD 3-Clause "New" or "Revised" LicenseBSD-3-Clause

Biometric Camera Sign in

An Python3 application to facilitate biometric sign in via an edge device with a camera. Developed on a Raspberry Pi 4.

This project came about from conversations with the Atlanta-based non-profit SafeHouse Outreach, dedicated to breaking the cycle of poverty, which is apparent on just about any walk through downtown (Peachtree Center) Atlanta.

One service SafeHouse provides is a meal several times a week, and each time a list of guests (i.e. individuals living on the streets) is required by the Georgia Department of Community Affairs Homeless Management Information System (HMIS). There are many repeat guests so in collaboration with SafeHouse, a biometric sign in device that could output list of guests (i.e. a meal log) was settled upon as a solution.

A Raspberry Pi 4 + camera was used for development, but future iterations could leverage more powerful hardware like a NVIDIA Jetson Nano.

Special thanks to SafeHouse for their support!

Installation

Instructions assume a default 'pi' user and sources cloned into the home directory:

  • Install Docker:

    curl -sSL https://get.docker.com | sh

    More info is available here.

  • Clone the repo to the home directory with the commands:

    cd /home/pi

git clone https://github.com/blakeflei/biometric_camera_signin.git

  • Create default configuration and download pretrained models:

    docker run \
      --rm \
      -v /home/pi/biometric_camera_signin:/home/biom/biometric_camera_signin  \
      -w /home/biom/biometric_camera_signin \
      blakeflei/arm32v7-biometric:20200508 \
      bash biometric_setup.sh

  • Start sign in app:

    cd /home/pi/biometric_camera_signin

bash start.sh

Building the docker image:

While the dockerfile is available for building in the docker folder, arm32 libraries for python and opencv-4.3.0 aren't available via pip or debian buster repos, so the build process requires compilation and takes several hours.

A docker image is available on docker hub and is recommended for use.

Dependencies

While the python code is platform independent, the docker image presumes arm32 architecture.

References

License

BSD 3-Clause License. Please see LICENSE file.