During her 4-month co-op term at our Data Analytics Centre, Nastaran Enshaei helped lay the groundwork to solve practical issues in gas metering. She is a fourth-year Ph.D. student in information systems engineering at Concordia University, and is already the accomplished author and co-author of many scholarly publications with a focus on artificial intelligence and deep learning systems.
Her main task was to write a scientific manuscript about accurate gas consumption monitoring. She was given the freedom to approach the project in her own unique way, while being supported and guided by the team. The goal was to develop a deep learning-based framework that takes real-time gas meter images and converts them to gas consumption.
Nastaran used a dataset containing more than 56,000 gas meter images, but many of them were blurry, noisy or shadowy due to variable lighting and weather conditions. She overcame the challenge by using data augmentation strategies, and the model achieved test accuracy above 95%.
The project will help consumers understand their natural gas usage by detecting leaks, reducing errors, flagging unusual patterns and providing other insights from monitoring data trends.
Working remotely, Nastaran integrated with the team and connected with colleagues through regular discussions. Even after returning to Concordia to continue her studies, Nastaran feels she can count on her colleagues for support and advice. “While I was only there temporarily, I wasn’t made to feel like that—I felt really involved in the group.”
Senior research officer Dr. Ashkan Ebadi says, “Nastaran is a gifted student! She was very dedicated to her work and to performing quality research.”