At Realeyes, we believe that we can assess the quality and relevance of video content by capturing and analysing an individual’s facial expressions and emotions as they watch it. Every day, we collect terabytes of video recordings across multiple countries, devices and web platforms, to determine how people feel while they watch video content. In this presentation, we will discuss how we use Spark and Databricks to enhance our data analytics and collection processes, and lean on predictive models that leverage facial expressions, behavioural cues and other metadata to predict business KPIs. We will also show how we combine Spark and Python libraries to build scalable validation and production pipelines in order to increase the robustness of our predictive models.
Javier is a Researcher and Data Scientist at Realeyes where he works researching and developing on predictive models combining his computer vision and machine learning expertise. Javier holds a PhD in Computer Science and Human Emotion Understanding from the Autonomous University of Barcelona. He joined Realeyes after 4 years of post-doctoral experience at the Imperial College London, where he was working on face analysis, emotion and deception recognition.