Real-time Audience Analysis

Firebase, Google Cloud, Node.js

Comment

Inspiration

As a public speaker one of the pain point is that I do not know about my audience

  • What my audiences FEEL during my speaking session
  • WHO are they?
  • What are they DEMOGRAPHICS

And the important is I want to improve my public speaking skill

What it does

Speaker+ allow speaker (or camera man) just take (audiences) photo during a speaking session, then audience stats will show up on the speaker dashboard in real time

How I built it

  • Progressive Web Application
  • Firebase (Firestore / Cloud Storage / Authentication / Cloud Functions)
  • Google Cloud Platform
  • Vision API

Challenges I ran into

  • Camera feature on PWA is limited
  • PWA in fullscreen mode is not supported on Chrome (v.61), But works fine on Chrome Dev (v.64)
  • The accuracy of Vision API

Accomplishments that I'm proud of

  • Audience demographics (gender / age) from analyzer comparing to face is 95% accuracy
  • Audience emotions (happy) is working

What I learned

  • Photo quality is major impact to accuracy (resolution, light condition, stabilization)

What's next for Speaker+

  • Detect audiences activities during the sessions (e.g. count number of hand-up, how many time they focus on their own mobile phone)
  • Make it supports multiple speakers
  • Make it supports audiences
  • Allow audiences to identify themselves
Share this project:

Updates