SelfieTV - Emotion Detection for TV Recommendation


It's Friday night and Sophie (25) and her friend Emma (28) sit down to watch TV with two glasses of Red Wine. As they relax on the sofa, they switch on their SkyTV - prompting them to take a selfie. Within seconds, Games of Thrones begins to play begins to play. They look at each other in astonishment......... how did it know?

Recommendation Algorithm

SelfieTV uses around 25,000 data points to recommend TV programs to users, based on age, gender, number of people and emotions.


NodeJS, Javascript, JSON, Jquery, Bootstrap


  1. API Integration
  2. Basing recommendations just on gender and age is always going to be stereotypical, but by combining with previous viewing history, location and social media trends it could be really predictive.

Accomplishments that we're proud of

  1. Emotion detection
  2. Integration with Sky data sets

What we learned

  1. When you combine new technologies in novel ways - you can deliver 10x improvements

What's next for SelfieTV

  1. SmartTV version
  2. Data Analysis for TV Producers
Share this project: