Exploratory Video Analytics

EXPLORATORY VIDEO ANALYTICS (2022)

• EVA is a visual data management system designed to make video analysis as intuitive as database queries. The platform supports a declarative SQL-like language and integrates a comprehensive range of computer vision models, enabling users to perform complex video analytics through simple query syntax.

• The following example demonstrates EVA’s capabilities by running an emotion detector on all detected faces within a video. Traditional approaches would require extensive boilerplate code for face extraction and emotion detection, but EVA simplifies this process into a single query:

  SELECT id, bbox, EmotionDetector(Crop(data, bbox)) 
  FROM MyVideo JOIN LATERAL UNNEST(FaceDetector(data)) AS Face(bbox, conf);

• This project was developed as part of the Georgia Tech Database Research Group over three semesters during my Master’s program, with primary contributions in machine learning implementation. Comprehensive documentation is available in the project documentation. Here is the research paper