A Convolutional Neural Network-Based Automated Approach for Confidence Analysis of Lecturers

A lecturer who exhibits confidence enables him to effectively deliver the lecture. Understanding a lecturer's confidence can provide insights into the conviction of their ideas, communication skills, and adaptability to questions. Confident lecturers can make a positive impact on the audience, leading to a more engaging and organized presentation. The Convolutional Neural Network (CNN), a Deep Neural Network (DNN) variation effective for classification problems, is utilized in this paper to determine the confidence level of the lecturers. The models were simultaneously validated using the split validation approach, which separates them into train, validation, and test and is then assessed using a confusion matrix. The trial findings demonstrated that the suggested technique produced an average accuracy of 96.47%. The proposed model correctly predicted the image with an average accuracy of 96% when three different confidence levels of lecturers were evaluated using confusion matrices. The CNN model was trained on a dataset of lecturer images categorized into three classes: high, medium, and low confidence levels. The proposed classification model technique has the potential to be highly beneficial in a wide range of lectures, offering a valuable tool for assessing and enhancing lecturer confidence, thereby improving the overall quality of the presentations.