日本語

• Disaster
• Visit demonstration page on HEROKU !
• Project background
• Features
• Requirements
• Usage
  • Use Docker
  • Other than using Docker
    • When trying to create numerical face data
• Component
  • 1. Web application
  • 2. The Application for creating numerical face data from camera
  • 3. Data synchronization between servers
• Project maintenance and contribution
• License

Disaster is a face search system for disasters using dlib, face_recognition, flask, PySimpleGUI and others, and what is written in Python.

How can you find a loved one in the event of immediately after a disaster? Disaster allows you to search for the face of the person you are looking for from the video taken by the camera installed in the shelter using Face Recognition technology.

Disaster is most privacy-friendly. The face information is replaced with irrecoverable numerical data, and even the developer cannot restore the original face image.

• Disaster web application demo
  
  

• The Disaster application window demo of creating 128D vector data from faces.
  

You can experience the Disaster web application. https://disaster-application.herokuapp.com/

Since I am using heroku's free plan, it takes about 30 seconds or more to start the web application.

It is an urgent task to find a solution to situation where safety of a loved one cannot be confirmed in a disaster. Currently, various companies are working to solve problems bringing together feasible technologies.

However, pre-registration before the disaster is often required, and/or family discussions are a prerequisite in advance. So it is expected that it will be difficult to confirm the safety of family after the disaster if some measures are not taken before the disaster.

If the situation when victim cannot be used smartphone, it is so difficult for family to try to contact victim immediately after the disaster.

These situations are occurring all over the world not only natural disasters but also wars or civil wars.

• Converts the image taken by a camera installed in the shelter into numerical data that cannot be restored to the original face image. Since privacy is of the utmost importance, the victim's face image is not displayed and immediately is removed.

• When upload a family photo to the Disaster web application, it will automatically look for similar faces and display information about when and in which shelter the victim was.

• Local governments, organizations and others can freely use Disaster.

• Ubuntu or similar Linux distribution
• NVIDIA GeForce GTX 1660 Ti +
  • Not needed if you just want to try Disaster. In that case, processing speed is extremely slow.
  • If you want to use GPU, you need to install the driver etc. by the method shown here.
• Python 3.7 +
• Docker 19.03 +
  • Only needed if you use Docker.
• Network camera or webcam
  • Not needed if you just want to try Disaster. In that case, use the attached mp4 file.

Execution environment example (My development environment)
Kernel	Linux 5.4.0-89-generic (x86_64)
Version	#100~18.04.1-Ubuntu SMP Wed Sep 29 10:59:42 UTC 2021
C Library	GNU C Library / (Ubuntu GLIBC 2.27-3ubuntu1.4) 2.27
Distribution	Ubuntu 18.04.6 LTS
Renderer	NVIDIA GeForce GTX 1660 Ti/PCIe/SSE2
Version	4.6.0 NVIDIA 470.63.01
AMD Ryzen 5 1400 Quad-Core Processor
Total Memory	16389096 KiB
Python 3.7.11(pyenv)

If you use Docker, please refer to here.
If you do not have the nvidia-docker2 package installed, the operating speed is extremely slow. However, if you just want to try Disaster, I think you can leave it as it is.
In that case, please use "Dockerfile".

If you want to run Disaster at normal processing speed but do not have nvidia-docker2 package installed, please do the following.

# For Ubuntu 18.04
distribution=$(. /etc/os-release;echo $ID$VERSION_ID) && \
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add - && \
curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list

sudo apt-get update
sudo apt-get install -y nvidia-docker2
sudo systemctl restart docker

For instructions on how to install the nvidia-docker2 package on each Linux distribution, see the official documentation installation guide.
https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html#installing-docker-ce

Also, use "Dockerfile_GPU" for the Dockerfile. In that case, change the file name from Dockerfile_GPU to Dockerfile.

Build python runtime environment and others are described here.

$ git clone https://github.com/yKesamaru/disaster.git

$ cd web_app
$ export FLASK_APP=main.py
$ flask run
 * Serving Flask app 'main.py' (lazy loading)
 * Environment: production
   WARNING: This is a development server. Do not use it in a production deployment.
   Use a production WSGI server instead.
 * Debug mode: off
 * Running on http://127.0.0.1:5000/ (Press CTRL+C to quit)

*** Access http://127.0.0.1:5000/ ***

You can use network cameras and USB cameras in Disaster.
Here, an mp4 file is used for the sake of simplicity.
Optional: Activate the Python virtual environment in advance.

$ cd create_face_data/shelter01
$ python create_face_data_app.py 
$ cd ../
$ python marge_npKnown.py

This system consists of three blocks.

Disaster
  ├ main.py
  ├ npKnown.npz
  └ others

People who want to find a family upload a family photo. It converts to non-recoverable numerical data on the server and searches for similar faces in the footage taken by the shelter's camera. Date, time and the shelter's name are displayed on the smartphone.

Disaster
└ create_face_data
    ├ marge_npKnown.py
    ├ npKnown_root.npz
    ├ shelter01
    │  ├ create_face_data_app.py
    │  ├ shelter01.mp4    
    │  └ npKnown.npz
    └ shelter02
        ├ create_face_data_app.py
        ├ shelter02.mp4    
        └ npKnown.npz

All the faces taken by the camera installed in the shelter are converted into numerical data.

The shelter01 and shelter02 directories were created to represent multiple shelters.
Click here for a more detailed explanation.

The face data file is shared between each server to prevent the entire system from going down.
This feature has not been created yet.

Disaster is maintenanced by Yoshitsugu Kesamaru.
I hope that similar projects will spread widely.

Since my mother tongue is not English, I'm not very good at it, so I would appreciate it if you could point out any description my using incorrect English.

Thankyou

MIT license