ATTENDANCE SYSTEM USING ONE-SHOT LEARNING

Abstract

The conventional methods for attendance tracking often involve manual processes, such as paper based sign-in sheets or biometric systems, which are time-consuming, error-prone, and cumbersome. With the advancements in computer vision and machine learning, the use of one shot learning techniques in attendance systems has gained significant attention. This thesis presents a comprehensive study on the design, development, and evaluation of an attendance system based on one-shot learning, aiming to provide an efficient and reliable solution for attendance management. In any organization, attendance is one of the most important things whether it can be an educational institute, or any corporate company. In most of the educational institutes, attendance of the students are marked manually, which is a very time consuming process if the number of students is huge. Maintaining attendance manually in the form of paper sheets is a very difficult task. There can be an error in filling records while entering data manually. These records can be easily manipulated. To solve this problem, we should have some computer aided solution to automate the attendance taking process which should be more reliable. Machine learning techniques are almost being used to solve most of the real world problems. Convolutional neural networks have become extensively employed in image processing tasks such as image classification, image segmentation, and object recognition and detection. These networks possess the capability to extracting characteristics or attributes from the input data. Here, we are using one shot learning, where only one sample of each class is fed to the model for training and later on this model is able to predict the unseen samples. Attendance marking systems play a crucial role in institutions, colleges, and corporations of all sizes. Through remarkable advancements in image processing, we have developed a system that capitalizes on this technology to simplify the process of managing attendance. Compared to other biometric authentication methods, face recognition has gained widespread popularity due to its convenient, non-intrusive, and contactless approach. The primary goal of this system is to identify faces and instantly recognize them by comparing them with the data stored in the database, ultimately recording attendance. By doing so, the system aims to streamline the arduous manual attendance process and improve its overall efficiency. Additionally, this system effectively tackles authentication and proxy-related challenges, as biometrics, including facial features used in face recognition, are unique and cannot be replicated. The system has been designed using OpenCV, dlib, Face Recognition libraries, and One-Shot Learning techniques for precise face detection and recognition. Notably, only one image per individual in the database is required, resulting in a space-optimized solution compared to traditional training-testing models. Attendance assessment is considered essential within the classroom environment in many institutions. Every institute, college, and organization has an attendance marking system. The evaluation of a student's performance is significantly influenced by attending lectures. As a result, our educational system places such a high priority on attendance. It is a valuable metric for assessing a student's performance. Therefore, we suggest a real-time attendance system that detects students by their faces after first identifying them. This model utilizes image processing, identifying the person and facial feature detection. Face Recognition is becoming more preferred over other biometric identification methods due to its simple, non-intrusive, and contactless methodology. The main objective of the system is to detect faces in real-time, recognize them, compare them with the database information, and record attendance. This serves to enhance and simplify the cumbersome manual attendance process. By leveraging the uniqueness of biometrics, specifically facial traits used in Face Recognition, the system effectively addresses authentication and proxy-related concerns. Unlike conventional training-testing models, to achieve accurate face detection and recognition, the system was created by incorporating OpenCV, dlib, Face Recognition libraries, and employing One-Shot Learning techniques.