In December, College of Computing and Informatics (CCI) adjunct professor Stephen Welch hosted an inaugural poster presentation of his students’ work in the atrium of UNC Charlotte’s Center City building. In addition to students and faculty, Welch invited industry partners and members of Charlotte’s tech community to attend and interact.
Welch, also VP of Data Science at Mariner, sees events like this as a way to, “...chip away at a real problem we have in Charlotte, which is the need for deeper and more impactful connections between academia and industry.”
Welch has found that facilitating collaboration between his students and industry professionals keeps his students focused and motivated. “It’s one thing," he said, "for students to hear me tell them in lecture that their work can be important, impactful, and relevant - but quite another when they really see that they can use what they learn in class to solve real world problems and make a difference.”
On the first day of his Computer Vision class, Welch made the following deal with his students: There would be no final exam, as long as each of them solved a real-world computer vision problem. “The showcase of that arrangement,” he said, “is the poster-presentation event. It gives students a chance to show off their work, and offers the broader community a taste of the cool stuff happening at UNC Charlotte.”
Attendee Phillipe Loher, Director of Machine Learning at Thomas Jefferson University, said, “It was inspiring to watch undergraduate and graduate students work together using Vision AI technology to make improvements in a field (molecular biology) beyond their immediate expertise. The multidisciplinary and project-based nature of the course, including the collaborative poster session, is an exemplary educational model that allows students to not only learn the latest AI technology, but the nuances needed to apply them to real-world scenarios.”
Greg Baker, Director of Research and Development at 2ULaundry, believes the class and event can prove immediately beneficial. “I'm extremely impressed with the results the students were able to achieve,” he said. “The accuracy is incredible and something that could have an immediate impact on our business.”
Looking forward, Welch may change the event’s format to include judging to identify the best presentations, but he did acknowledge there was a crowd favorite.
Graduate student, Anbumalar Saravanan (pictured), who worked independently while most students worked in groups, developed “Combat Finder” to detect and battle bullying in schools. Saravanan created a dataset of images depicting common bullying actions, such as kicking and punching, to create a computer vision algorithm to detect bullying in video streams captured by the increasing number of security cameras in schools.
"I was overwhelmed by the positive reaction," said Saravanan. "I am a mother of a three-year-old, which made it difficult to work in a group. I knew completing the project would be a challenge, but the positive reactions are a reward that still keeps me going. I even received valuable suggestions from several people about metrics that would improve my model and ways Combat Finder could be used in other places like psychiatric wards or jails to prevent chaos. This is the kind of exposure that never would have come in a regular classroom presentation setting."
“I’m really proud of the effort our students and TAs put in to make the projects and the event a success. We did our best to make sure that each group had a well prepared 3-5 minute pitch that was accessible to non-technical experts,” Welch said. “This really paid off. The room was alive with engaging conversations.” Welch is already planning a reprisal of the event in 2020. “Bigger and better,” he said. “We would definitely love to have more involvement from the community and maybe even a small budget for snacks so we don’t have to make them at home.”