Students at UH Mānoa’s College of Engineering don’t just study theory and concepts – they roll up their sleeves and attack current challenges within our community. More and more, the college is pivoting to infuse industry-integrated courses into their four-year curricula, where, through the course of a semester, students are paired with local organizations to design and develop solutions to address actual problems.
Software development for enhanced check-in process
One such opportunity, the Community Innovation Mentorship Program (CIMP), was born through a collaboration between local company DataHouse, Transform Hawaiʻi Government, Hawaiʻi Technology Development Corp (HTDC), the State of Hawaiʻi Department of Agriculture, and the College of Engineering.
The program, under the umbrella of the TRUE Initiative, afforded a team of seven undergraduate computer engineering students the opportunity to tackle software development projects side by side with industry experts. The team started by utilizing tools and techniques to identify and refine problems, to which they then brainstormed and prioritized innovative solutions involving programming frameworks and methodologies, user authentication, technical architectural and database design, middle-tier business logic, and front end UI/UX frameworks.
That client was the Hawaiʻi State Department of Agriculture’s Animal Quarantine (AQ) unit, and the problem at hand was one that their inspectors, pet caretakers, and the many pet owners experienced on a daily basis: the inefficiencies of the animal check-in process. Over the course of five months, the CIMP Team developed a new check-in process and queuing method for the Animal Quarantine Airport Station, featuring two large monitors and an iPad-driven kiosk check-in station that allows the AQ staff to utilize electronic devices to visually guide the processing of pets and pet owners.
Their solution also streamlined the staff’s workflow with a new messaging system, providing a quicker means of communication between front inspectors and the back pet caretakers, and helped reduce pet owner anxiety by providing up-to-date information on their pet’s status via an accompanying app. “It was encouraging to watch the successes and progress from the hard work of the UH students and mentors. The Community Innovation Mentorship Program works very well.” said Dr Isaac Maeda, Administrator of the Animal Industry Division, which also includes the title of State Veterinarian of the Hawaiʻi Department of Agriculture (HDOA). The program concluded with a final team presentation and certificate ceremony at the Entrepreneurs Sandbox in Kakaʻako on May 18.
Machine learning for dental radiograph classification
At the graduate level, Dr. Il Yong Chun’s spring course, Computational Image Processing and Computer Vision, afforded students from various engineering and computer science backgrounds the opportunity to understand industrial demands and challenges found beyond the textbook through a flipped-classroom approach, culminating in a project with Hawaiʻi Dental Service (HDS).
Students focused on areas such as image restoration, image segmentation, and pattern recognition, using algorithms, statistical modeling, and artificial intelligence (AI) to achieve their outcomes.
“EE616 targets graduate students who want to understand the mathematical foundations underlying computational data science solutions in image processing and computer vision and then eventually develop new solutions,” said Chun.
In a modified version of the flipped classroom approach, Chun provided students with partial lecture notes in advance of the class, encouraging more student engagement during the allotted class time.
The final project, in partnership with Hawaiʻi Dental Service (HDS), put to use the students’ enhanced understanding of image pattern recognition and machine learning, allowing them to develop, train, and test AI systems that can identify dental radiographs by those that do and do not indicate dental disease. Such software could prove helpful in assisting dental insurers to identify inaccuracies in insurance reporting, among other applications.
In the end, the data from HDS was not suitable for the project at hand, so students utilized a publicly available dataset to develop AI models that would accurately classify the radiographs. Amongst the results of the seven class members, the most accurate solution performed the task correctly 87.5% of the time.