This research paper presents a Classroom Activity Detection system that addresses the challenges posed by the transition to online and hybrid learning models, as well as the increasing number of students in physical lectures. The system leverages machine learning and computer vision techniques to monitor and categorize student activities in real time, allowing teachers to ensure engagement and focus during lectures. The system is designed to work in both instructor-present and instructor-absent scenarios, eliminating the need for strict invigilation during exams. Implementing TinyML, a field of machine learning focused on deploying models trained on MobileNetV2 architecture on resource-constrained devices, makes the system portable and practical for various educational settings. It is deployed on Raspberry Pi platform, which offers affordability, ease of use, and versatility without compromising processing power. This innovative solution addresses the critical challenges faced in modern classrooms and enhances the teaching–learning experience.