Abstract: Student activity recognition and classification is an emerging application of machine learning in the educational domain. This project aims to identify and classify various student activities—such as reading, writing, using a mobile phone, or sleeping—based on image or video data. A convolutional neural network (CNN) is employed to extract spatial features and learn activity-specific patterns from labeled image frames. The backend is built using FastAPI for efficient model deployment and API integration. A React and Tailwind-based frontend allows real-time interaction and visualization. The system supports data augmentation techniques to improve accuracy on limited datasets. Activities are preprocessed and labeled into a CSV for model training and evaluation. The model’s performance is validated using metrics like accuracy, precision, and recall. The goal is to assist educators in monitoring classroom behavior and enhancing learning outcomes. This solution can be extended to real-time surveillance or attendance systems.

Keywords: The main keywords of the project are Identifying and analyzing actions performed by students using visual data, Applying algorithms that allow systems to learn patterns and make predictions from data, A deep learning model used for image-based activity classification. Categorizing images into predefined labels based on their visual content.

| DOI: 10.17148/IARJSET.2025.125261