Hari Santhanam: synthesizing computer vision data

Friday, Sep 25, 2020
by Sharon Adarlo

Hari Santhanam, Class of 2020


Santhanam majored in electrical engineering and earned two undergraduate certificates: Statistics and Machine Learning from the Center for Statistics and Machine Learning (CSML), and Robotics and Intelligent Systems.


Santhanam’s independent project for his CSML certificate, “Video Synthesis: Binary Masks to Frames via DeepInversion,” delved into machine learning and computer vision applications. Specifically, his senior thesis, dealt with synthesizing video from a pre-trained neural network.

“Machine learning models rely on data for training, so they can help make real-world predictions. The acquisition of such training data can be arduous in certain situations,” said Santhanam. “For example, some models are developed and trained using data that is privacy protected. As a result, such datasets become inaccessible to researchers seeking to train new models and make predictions. If we can synthesize ``training data’’ from a pre-trained model, this can greatly aid potential knowledge transfer.”

A practical application of this would be a system that uses federated learning, a technique that allows programs to train and improve from decentralized data. Hari said that a system that uses federated learning can benefit from his project because it would eliminate the need to transfer data between servers, thus reducing computational time.

For his project, Santhanam used a new technique called DeepInversion to synthesize video training data from the DAVIS or the Densely Annotated Video Segmentation dataset, which is made up of 50 video segments of cars, planes, people and various objects.

Santhanam became interested in pursuing this research study after working on machine learning with his thesis advisor, Niraj Jha, professor of electrical engineering.

“I became intrigued with machine learning because it can be applied to many different fields of study,” he said.

Before delving into computer vision and machine learning, Santhanam was an intern at the Princeton Plasma Physics Laboratory, where he studied oscillations in plasma. He also did an internship at Lightening Energy, a New Jersey-based company that develops advanced batteries and technologies to improve the reliability of the power grid.

After he graduated, Santhanam enrolled in the University of Pennsylvania’s master’s degree program in robotics, concentrating on computer vision and machine learning. After he earns his master’s degree, Santhanam eventually wants to work as a researcher in industry and pursue a doctoral degree.

Extracurricular Activities:

Santhanam was a clarinet player for the Princeton Wind Ensemble and played club tennis. He was also a head undergraduate tutor at the McGraw Center, where he specialized in introductory engineering courses.

For Fun:

Santhanam enjoys playing ping pong, watching Netflix, playing sports, and going to the gym.