Computer vision emotion 😜 detection game with TensorFlow backend.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
partypi
.dockerignore
.gitignore
.travis.yml
Dockerfile
LICENSE
MANIFEST.in
Pipfile.lock
Procfile
README.rst
app.json
docker-compose.yml
google6af5bff2c16251ad.html
gunicorn.conf
heroku.yml
requirements.txt
runtime.txt
screenshot.png
setup.cfg
setup.py

README.rst

Party Pi

license  

Emotion detection game using computer vision. Note: This repo is no longer maintained. Contact @justinshenk if you have questions about running it.

Description

Interactive game that ranks players by their ability to show an emotion (eg, "Show surprise").

Emotion detection is accomplished using an Inception-based neural network trained in TensorFlow on the FER2013 dataset. Face detection is accomplished using OpenCV's Haar cascade.

Demo

Pics from the first demo at the Institute of Cognitive Science 2016 Christmas Party

screenshot of game

Play

Visit https://partypi.net with Desktop Chrome or Firefox.

Run locally

Clone repository:

git clone https://github.com/JustinShenk/party-pi.git
cd party-pi

Install dependencies:

pip3 install -r requirements.txt

If using Ubuntu, install tkinter with sudo apt-get install python3-tk

gunicorn -c gunicorn.conf main:app

Open a browser to https://localhost:5000.

Docker

Pull the docker image with docker pull justinshenk/partypi:latest.

Run the server interactively with docker run -it -p 5000:5000 partypi and open a browser to https://localhost:5000.

In Chrome (or Firefox), click Advanced and proceed past the warning.

Raspberry Pi

Additional (optional) arguments: --picam for piCamera module and --slow to slow down the countdown.

Select Easy or Hard Mode (left or right arrow keys).

Awards, Press, Demos

  • Featured in the Intel Developer Mesh Editor's Picks (August 2017)
  • Awarded the 2017 Intel Hack Challenge prize
  • The technology was demonstrated at Intel's AI booth at NIPS 2017
  • Party Pi was demoed for students and faculty at the 2016 and 2017 Osnabrück University Institute of Cognitive Science Christmas parties
  • Received the 2018 Open Source Software Prize from Intevation GmbH

TODO

  • Fix alignment of user interface elements
  • Fix display of elements
  • Fix mobile image processing
  • Test latest version on Raspberry Pi (support is deprecated)

Author

Justin Shenk (@JustinShenk) created Party Pi.

License

MIT