An impressive finish on a computer vision conference challenge has paved the way for a computer science student’s first trip overseas.
Tina Tran, an accelerated bachelors to master’s student and undergraduate researcher at the Center for Research in Computer Vision (CRCV), earned a trip to Milan to present her research courtesy of CRCV founder and director Mubarak Shah. Shah generously offered the prize as motivation to pursue and complete a workshop challenge sponsored by Google DeepMind at the European Conference on Computer Vision (ECCV) held last month.
Tran has known she wanted to pursue research since she was an elementary school student, a passion sparked from her participation in STEM fairs. She was drawn to UCF for its highly ranked computer science programs, its proximity to her family in Sarasota and the financial incentives offered.
“Since my family would be sending two kids off to college at the same time, I wanted to reduce their financial burden,” Tran says. “I was awarded scholarships that completely covered the cost of my education and housing to go to UCF. My twin sister was also covered by scholarships to come to UCF and we were both accepted into the Burnett Honors College.”
Tran began her academic journey at UCF with the intent to study machine learning. That changed when she learned of CRCV’s world prominence and decided to learn more about burgeoning research and advances in computer vision.
“I pursued an honors undergraduate thesis to build a foundation for AI research and boldly decided to invite Dr. Shah to my honors thesis committee,” Tran says. “After my successful thesis defense, Dr. Shah invited me to work with him. Everything seemed to fall into place after that, and with renewed confidence I joined Dr. Shah’s group this past summer semester.”
As a computer vision researcher, Tran uses algorithms to help computers “see” and interpret data, using artificial intelligence to train computers to complete visual tasks.
“I work in the visual language field to create and improve multimodal AI models that can learn from images, videos and text,” she says. “Generally, these models can display complex visual reasoning to answer questions such as, ‘Where is the dog in this video?’ or ‘Please summarize this video.’ The potential applications for this technology are vast, spanning many fields such as defense, security, medicine, education and robotics, to name a few.”
Although Tran has been pursuing research at CRCV just a few short months – she joined the lab this past summer – she decided to tackle a challenge at one of the world’s foremost computer vision conferences. She and CRCV doctoral student Joe Fioresi worked together to take on the Second Perception Test Challenge at ECCV. Tran says their category, the grounded video question-answering task, was the most challenging of the six options offered. Not only were researchers tasked with tracking an object, but also with teaching a model to answer a question to discern the correct object to track.
Shah encouraged his researchers by generously offering to send them to ECCV in Milan if they won the challenge. With their proposed solution at the top of the leaderboard for two months, a trip to Italy for the pair seemed imminent. In the final 15 minutes before the challenge deadline however, they were surpassed by another team that kept their score hidden.
“We were a little bit disappointed with second place,” Tran says. “The reason we kept our results public was because Joe and I wanted to be transparent about our results. I believe that research is about sharing what has been accomplished to drive other researchers and humanity to reach greater heights. I am proud of our entry.”
Shah was equally proud of their entry, and graciously sent his researchers to Milan as a reward for their efforts. It was a whirlwind experience for Tran, who presented her research to an international audience at the conference for the first time.
“Poster sessions, oral presentations and demos were happening parallel, so it was very overwhelming,” she says. “I was able to learn about the cutting-edge research happening in the field and what researchers all over the world are doing. What I enjoyed the most was the atmosphere. It was exciting, inspiring, and motivated me to continue pursuing research.”
Tran also had the opportunity to take in the sights of Milan, including its renowned cathedral Duomo di Milano, travel the city via the metro and enjoy classic Italian fare like gnocchi, risotto, gelato and tiramisu.
Though Tran planned to attend the conference on her own, her family had some reservations about sending their daughter off alone on her inaugural trip overseas. They came up with a novel solution to ease their concerns – her mom, dad and twin sister joined her in Milan for a long overdue family vacation. They spent two weeks preparing for the impromptu trip, rushing to renew passports, learning a bit of Italian through Duolingo and, for Tran’s sister Victoria, also a computer science student, developing a plan to stay on top coursework by arranging online sessions.
“I learned a lot from this experience and had the opportunity to present live to an international crowd in Italy,” Tran says. “I would like to thank Joe, Dr. Shah and my family for supporting, mentoring and making it possible for this amazing opportunity to happen.”
Story by Bel Huston