ABSTRACT: Outcome-based Education (OBE) emphasizes measurable learning outcomes aligned with practical skills, incorporating frameworks like the Revised Bloom’s Taxonomy (RBT) to cultivate higher-order thinking skills (HOTS) such as critical analysis and creativity. However, manually evaluating complex student assessments such as hazard identification in image tasks is time-consuming, subjective, and prone to inconsistencies, particularly with large student submissions. Traditional automated tools, like multiple-choice graders, fail to assess HOTS, leaving educators overwhelmed by administrative burdens. This study proposes a deep learning (DL) approach to automate the scoring of interactive student assessments within an OBE-RBT framework. Focusing on a boxed visual hazard identification activity, the research leverages DL models trained on 86 student submissions from a manufacturing engineering Occupational Safety and Health (OSH) course. Images were preprocessed, annotated for correct hazard identifications, and split into training (60%), validation (20%), and testing (20%) sets using Roboflow. The model achieved promising results, with precision and recall 97.7% and 98.8%, respectively, and a mean Average Precision (mAP50) of 99% after 200 epochs, demonstrating robust boxed hazard detection. By automating scoring, this approach reduces faculty workload, enhances grading accuracy, and aligns with OBE’s pedagogical goals. Unlike prior AI research focused on text or structured evaluations, this study innovatively targets HOTS within visual tasks, bridging a gap in educational assessment technology. The study’s findings imply that deep learning has the aptitude to fundamentally change assessment practices in education, thereby freeing up educators to allot more time to student mentorship, research, and extension activities, while simultaneously enabling the integration of artificial intelligence into pedagogical approaches.

Keywords: EduTech, object detection, learning assessment, Roboflow, OBE, RBT, deep learning