AN ADVANCED STUDENT MONITORING SYSTEM
FOR PHYSICAL CLASS ROOMS

Samridhi Dev and Tushar Patnaik have proposed a face recognition-based attendance system that is more accurate, efficient system. The system uses The system uses Haar classifiers, KNN, CNN, SVM algorithms to achieve high accuracy in face recognition, and has been tested under various conditions. The system is also non-invasive and easy to use, and requires minimal installation[1]. It's important to note that this system is primarily designed for attendance purposes and is not intended for classroom monitoring or assessing attentiveness levels.

Weiqing Wang and Kunliang Xu's research focused on classroom emotion detection. They designed an algorithm that uses MTCNN for face detection and image segmentation, creating a database for visual emotion classification. MTCNN outperformed VGG16 and ResNet18 with an accuracy of 93.53% in recognizing students' facial emotions[2]. It's worth noting that this system, like the previous one, is not designed for use in physical classrooms. It also lacks the ability to track multiple students simultaneously and cannot detect behaviors, setting it apart from our proposed system.

Recent research highlights machine learning's effectiveness in measuring student engagement through methods like keystroke analysis for programming activity and facial expression and body language analysis for presentations. These techniques, including eye-tracking and EEG, prove applicable in both face-to-face and online learning settings[3]. It's important to note that this system is specifically designed for online environments and is not intended for physical classrooms. Unlike our proposed system, it cannot detect multiple students simultaneously.

Jialu Wang et al. explored video super-resolution through generative adversarial networks (GAN) and edge enhancement in 2021[4]. Their model, applied to low-resolution and blur videos such as closed-circuit television, achieved realistic and detailed results through adversarial loss. The edge enhancement function, utilizing Laplacian edge module, further improved outcomes. The inclusion of perceptual loss enhanced visual experience. The approach was effective across datasets, with notable benefits observed on the Vid4 dataset and other low-resolution videos. In our proposed research we use the more improved and advanced of GAN model which is known as ESRGAN [10] model for better enhancement to detect students from video.