Building Skills, Confidence, and Community at the 2024 Student Research Showcase

Is that tasty-looking morsel on the forest floor an oyster mushroom, or a chantarelle? Yunxuan “Kaya” Rao, Master’s student at Northeastern Seattle, knows how to find out.  

Rao was one of 34 students who presented posters at the Khoury Fall Master of Computer Science Showcase in December, on research they completed under 8 different faculty mentors over the course of the semester. The topics ranged widely – from real-time environmental monitoring, to high-tech home organization, to the detection and translation of ASL using computer vision – but for Dennis Chenard, Rao’s convolutional neural network (CNN) to identify 100 different varieties of mushrooms was particularly impressive. 

“The poster was well presented. There were some late data sets and image additions that were taped to the poster board, the level of commitment to improve on and best represent the research results at the showcase was impressive (and something our judges were also impressed by),” said Chenard, the Director of Industry Partnerships for Student Learning. “Framing the hard work into a succinct poster board is a challenge; being able to speak about their work to any random person that comes by…(is) going to help them in their future.”  

“I really feel more confident about myself, more confident about my computer science skills, knowing that other people value what (I made),” Rao said. “I didn’t know if people would like my project or not. I thought I would be just standing there, and maybe talking to my classmates; I didn’t know people would actually come up to me because they are interested in this topic.”

Rao’s mushroom identification network grew out of her work in Professor Bruce Maxwell’s computer vision and pattern recognition class – one of the first AI-related classes she’d ever taken. Rao joined NU Seattle as an Align student in the Fall of 2022, and her background was in physics; most of her Khoury classes thus far had focused on web development.

“Before this semester, I wasn’t interested in doing anything in AI,” Rao said. “I really feel like I have done a lot (in this class). I also have put, like, so much time in, but it’s a lot of fun and the professor’s very responsive and supportive…I would say it’s the best class I’ve ever had.”

So when it came time to pick a topic for her final project, Rao wanted to take her newfound interest in deep learning further. She also found herself inspired by the area around Northeastern’s Seattle campus.

“I always want to try mushroom picking, but I was never able to do so because I can only recognize those mushrooms that they sell in the grocery store. Everything outside of the grocery store, I don’t know if it’s edible – I don’t know what kind of mushroom it is,” she said. “It also seems pretty fun, because this is Seattle – there’s rain forest in this area.”

Everything went smoothly at first, but Rao hit a major hurdle when she tried to scale her model up from a proof-of-concept identifying a handful of different mushrooms, to identifying 100 different varieties. The model was getting more accurate as she trained it, but it was improving way too slowly to be ready in time for the showcase. Rao had to get creative, switching the pre-trained network she was building on top of from MobileNet to DenseNet, and using data augmentation like RandomFlip, Random- Rotation, and RandomZoom to ensure her data set was diverse enough to train the new model.

On the day before the showcase, Rao’s mushroom identification network was still only achieving about 75 percent accuracy. Rao was nervous, and even considered cutting down the number of mushroom varieties her model could identify. At a friend’s suggestion, she instead unfroze the CNN’s last conventional block so she would have more parameters to train on, and set her model up to train overnight.

And in the morning, Rao woke up to a model returning nearly 84 percent accuracy.

Delighted, Rao taped the updated results to her poster and took off for the showcase, where her project was awarded first place by the event’s panel of secret faculty and alumni judges.  

“It was an amazing experience. It helped me realize that I can do more than I thought I can do,” Rao said. ”I finally feel like the physics knowledge I got in my undergrad and all the computer science skills I got in this program just sort of connected. It definitely helped me to learn things in this class and in this project.”  

Chenard, for his part, was excited to get a peek inside the unique research opportunities baked into Northeastern Seattle’s courses.  

”Research takes all kinds of different forms, right? There’s significant research happening in these classes,” Chenard said. “The professors worked with the students on framing the research elements of those course projects into these posters, which I was super impressed by; they were able to really show the research angle as it relates to the projects they were tackling.”  

As much as he enjoyed seeing it for himself, it’s even more important to Chenard that Northeastern’s alumni and industry partners get a chance to see the creative, innovative ideas that Northeastern Seattle’s students are testing out.  

”We want them to be not just passive participants in research, but active in the sense of providing the projects and being there as part of the process, as mentors and as the clients, and then ultimately (as) the benefactors of the research and the talent development,” Chenard said. ”I find when we go back into these academic environments and we’re in the jobs that the students are dreaming to be in one day, it reignites your spark for what you do.”  

Both Chenard and Rao hoped that the showcase would encourage others in the Northeastern Seattle community to get curious about the opportunities to complete and share research that the campus has to offer.  

 “You don’t have to worry about if it’s just a class project, if the scope is too small; just apply for it first. You might get invited, and people might be interested in your project,” Rao said. “In Northeastern University, in this program, I have (felt) my confidence gradually building up…(and) I think (other students should) have more confidence in what you are doing, more confidence in your work.”