the-politician-behind-the-poster

Inspiration

More than 45% of voters don't know any of the lead candidates for the upcoming european election. We want to increase transparency in politics and enable people to inform themselves more easily.

What it does

Our application is runnable on any mobile device, that can scan a campaign poster and provide a summarized overview on the candidates main topics, key contributions and opinions as well as a brief overview about the candidates party.

How we built it

We used Google Vision AI to scan the image and Google Natural Language Processing API in order to recognize and match the politician from the campaign poster with their profile. The topic overview is generated by an Author-Topic-Model that was trained on all recent plenary sessions for this election period. The session with multiple politicians participating is modeled as a distribution over topics that is a mixture of the distributions associated with the politician. Our overall tech stack contains a Python Backend (Flask), Machine Learning Model (Author Topic Model) and Frontend (React, ChartJs, WebcamJs,) and several API's (Google AI).

Challenges we ran into

States in React in Ecma Script 6 and Training Time for the actual Model (it would take several days, so we only used a small sample of our data set).

Accomplishments that we're proud of

We are really proud of our base concept as well as the ideation phase. It is incredible how team work and being focused on the same challenge can be fulfilling and how much we could accomplish in just 22 hours.