Search for: All records

This research full paper presents research around the Engineering Projects in Community Service (EPICS) program that serves two key purposes to: 1) provide a structured approach for engineering students to engage in real-world, service-based projects and 2) provide technical support and expertise that may be critical to local and global community organizations. Hence, EPICS strives to offer a platform that fosters the collaboration of engineering students and communities. EPICS helps develop undergraduate students’ professional skills extending beyond the theoretical knowledge acquired in classrooms. EPICS has been a fixture in engineering education for over 15 years, with a strong focus on curricular and pedagogical interventions to help students gain professional skills. The purpose of this paper is to explore the perspectives of over 650 students who participated in EPICS at a U.S. university during the academic years of 2019/2020 and 2020/2021. We used natural language processing (NLP) to thematically analyze students’ responses to an open-ended survey administered at the end of their semester participating in the EPICS program. Students’ responses reflect their perspectives on the design process, teamwork, real-world experiences, and the challenges they face during the design process related to other people and the program. In our findings, students’ least favorite parts of EPICS were lectures and design reviews, while their favorite parts of EPICS were teamwork and engaging with community partners. Understanding the themes emerging from the data can help us better implement community-based educational initiatives and find ways to better engage students in community service-learning projects. Our research provides implications for practice and research. 

Model calibration measures the agreement between the predicted probability estimates and the true correctness likelihood. Proper model calibration is vital for high-risk applications. Unfortunately, modern deep neural networks are poorly calibrated, compromising trustworthiness and reliability. Medical image segmentation particularly suffers from this due to the natural uncertainty of tissue boundaries. This is exasperated by their loss functions, which favor overconfidence in the majority classes. We address these challenges with DOMINO, a domain-aware model calibration method that leverages the semantic confusability and hierarchical similarity between class labels. Our experiments demonstrate that our DOMINO-calibrated deep neural networks outperform non-calibrated models and state-of-the-art morphometric methods in head image segmentation. Our results show that our method can consistently achieve better calibration, higher accuracy, and faster inference times than these methods, especially on rarer classes. This performance is attributed to our domain-aware regularization to inform semantic model calibration. These findings show the importance of semantic ties between class labels in building confidence in deep learning models. The framework has the potential to improve the trustworthiness and reliability of generic medical image segmentation models. The code for this article is available at: https://github.com/lab-smile/DOMINO.