CS Professor Granted NSF CAREER Award for Work in Video Databases

Joy Arulraj, an assistant professor in the School of Computer Science (SCS), was awarded a National Science Foundation (NSF) CAREER award for his work in video database systems.

Arulraj’s research focuses on designing database systems for analyzing unstructured data, including photos, videos, or podcasts, using deep learning models to discover previously unseen connections and insights.

Arulraj says that while traditional database systems are not capable of querying unstructured data, advances over the last decade in computer vision have made it possible to develop an artificial intelligence (AI) relational database system for automatically analyzing videos at scale and extracting new insights from them.

Until now, the focus of researchers in this field has been to increase the accuracy of vision models as much as possible. The issue now, Arulraj says, is that current deep learning models for computer vision have become computationally too costly. They also often require a user to be familiar with low-level programming. His research aims to reduce the computational cost of extracting new insights and make it easier for users to query unstructured data.

“Previously, video or audio analysis in diverse domains was done manually by humans and this was a very tedious process,” Arulraj said. “But now we can use computers to do the tedious part and let humans do more creative things on top of these insights.”

The outcome of Arulraj’s latest research is the open-sourced EVA AI-relational database system that supports deep learning models. EVA is beneficial in several fields. For instance, EVA can help analyze a city’s traffic videos to pinpoint factors that lead to accidents. Another way it could prove helpful is through monitoring social media. Videos can be hard to moderate manually, but Arulraj said that EVA could facilitate automated analysis of those types of content.

The CAREER grant will help support the development of the EVA system. Arulraj said that he and his students hope to improve the capabilities of EVA to support large language models, like ChatGPT, and to improve its speed using novel AI-centric query processing algorithms.

“These days, with all the excitement around language models and image generators and stuff like that, it’s exciting to see the possibilities of what future applications would look like with the EVA AI-relational database system.”

Along with the CAREER Award, Arulraj was also recognized for his work on video database systems by being named an IEEE TCDE Rising Star in 2022.

"This is a well deserved recognition for Joy. It has been great to see him develop this exciting research agenda on AI-relational database systems from scratch after starting at Georgia Tech -- it also dovetails perfectly with our school's new strategic focus on Foundations of AI," said Vivek Sarkar, SCS Chair and the Stephen Fleming Chair for Telecommunications in the College of Computing.