|
|
|
The performance of such attendance marking systems is based on how accurately the users are authenticated and they are measured using the true positive rate, true negative rate, false positive rate, and false negative rates. The true positive rate of the proposed system is 91.66% (14% greater than that found in existing literature). The tests conducted accounted for varying environmental setups ranging from illumination changes(Outdoors, Indoors, partially outdoors) to presence of occluding accessories glasses, cap, with dyed hair etc. Even cases where the user may attempt to proxy another user’s face with a picture were added in order to raise the robustness of the system. For faces that are foreign to the system, the false positive rate is 5% which is 9% less than that of the state of art, making the system more secure. The false positive rate for the users of the system is 9.52% while it is 28% for some the existing systems. The false-negativity rate was found to be 8.33%, making the system architecture more reliable as the registered employees will not be rejected wrongfully. The true negative rate is 90.47%, which implies the system will detect the non-users of the system correctly. The overall system time complexity is also improved compared to the literature as well the recognition rate and its instant updation to the main database takes approximately 5 seconds. The proposed efficient architecture also reduces the risk of a data breach and mitigates a wide range of security issues. It has an improved face recognition runtime than existing systems where even if a separate microprocessor is used which takes time to load, the weights of the model and process the recognition module. Here, the recognized individual id is directly fetched from the database. The proposed architecture, despite using a Deep CNN requires less computational power, is less time consuming for the user and is capable of handling a large data without compromising its performance, making it suitable for use in a large organization, unlike the existing systems in the literature considering the number of students, employees, teaching and non-teaching faculty, and departments. |
