Inspiration

  • In my graduate school era, I researched about elderly persons lifestyle in 2 years. Then I found out they are sometimes in trouble about fraud by high pressure sales person.
  • Now I'm working as sales person, that's why I understood sales person's thinking and feeling. I know all sales person do try and error to sell product, and get growing to be a skillful sales person. Also bad sales person has growing to sell fraudly high expensive product.
  • Elderly persons is too hard to decline sales persons's talk, they need help to protect from high pressure sales person. To answer their needs, I developed door phone to protect from high pressure sales person.

What it does

This product is a door phone system to protect the elderly persons from high pressure salesman. This dppr bell learn visited person face, and if the high pressure salesman or disliked person come home and ring the bell, this system can detect sales person and to notify danger or ignoring the salesman.

How I built it

  • Door bell hard ware is Raspberry PI.
  • Face detector is OpenCV.
  • Face verify function is used in Azure Face API.
  • Sound recognition is working on Watson, Speetch to text.
  • Two way communication system is to built by Node.js, Express, SocketI.O.
  • Backend server is created by Ruby on Rails.
  • NodeJS app and Rails app is working on CLOUD FOUNDRY.
  • DB is working on Compose.
  • WAV file and IMG file is saved in S3.
  • Dashbord and door bell system frontend is created by VueJS.

Challenges I ran into

  • First time to use IBM Blue Mix. I was in trouble deploy app.
  • It is too short time and resources to build DNN model. I substitute it to Azure Face API.
  • same time to work two way communication and face detect is so difficult.

Accomplishments that I'm proud of

  • Built two way communication system by full scratch
  • App is successfully deployed on the IBM BlueMix

What I learned

  • two way communication system is difficult to create.
  • PaaS service is helpful to us to crate powerful machine learning system.

What's next for sales pause

  • build DNN face detector without using PaaS service.
  • create same concept app for B to B. Many retails or restaurants are in trouble for so many commint and telling sales man, because sales man always come in busy time. I have a solution to protect retails and restaurants. That's to create Pepper system, to detect sales person face, and say "Get out here!!".
Share this project:
×

Updates