YOLOv8 in Education: Automated Evaluation of Teacher Presentation Skills

The assessment of teacher effectiveness is pivotal in maintaining and elevating the standard of education.Traditional evaluation methods often rely on subjective observations, making it challenging to maintain consistency and objectivity.To address these challenges, a novel Teacher Effectiveness Evaluation Assistant System (TEAS) is proposed that leverages the YOLOv8 object detection technique to objectively assess a teacher's performance based on their presentation content.To achieve this, we gather lecture videos and deploy YOLOv8 algorithm to train the data for the presence of two components in their presentation namely "Animations and Architectures".The dataset consists of annotated frames from these videos, where key visual elements like animations and architectural diagrams were labeled for detection.This dataset was utilized to train and evaluate the YOLOv8 model and compare its performance against other models.The trained YOLOv8 model was tested on test data to detect the components and thus assisting in evaluating a teacher.The proposed strategy generates a presentation score by assessing the density of animations and the extent to which architectural diagrams are covered, providing a measurable criterion for evaluation.The results obtained showed that YOLOv8 performed better in comparison to other models namely Faster regions with convolutional network model (Faster RCNN), SSD (Single Shot Detector), YOLOv5 and YOLOv7 models.The YOLOv8 model achieved an accuracy of 92.05% which was 27.05%, 36.35%,12.1%, 8.2% higher than Faster RCNN, SSD, YOLOv5, YOLOv7 models respectively.Therefore, YOLOv8 model can effectively detect component's presence in the presentation content of teachers, thus assisting in evaluating the instructors.