The research study, “Human Identification System Based on Ear Shape Using Convolutional Neural Networks”, aims to provide a secure and contactless biometric alternative for human identification. Utilizing the EarVN1.0 dataset, the study trained five models: MobileNetV3-Large, VGGNet-19, ResNet50V2, EfficientNetV2-S, and a modified AlexNet, achieving accuracies of 79%, 22%, 79%, 91%, and 52%, respectively. Through hyperparameter tuning of the EfficientNetV2-S model—focusing on batch size, pooling layer, dropout, and maximum epoch limit—the accuracy was improved to 92%. The research was conducted in Python, with a GUI implemented using Python Tkinter, which is compatible with macOS and Microsoft Windows. This project was developed by me and my team as part of our thesis research. The source code can be accessed on my Github. This research has been published on ScienceDirect and can be accessed here.
GitHub - nadyatyandra/Human-Identification-System-based-on-Ear-Shape-using-CNN
In this project, I took the role of a Machine Learning Engineer. Some of my responsibilities included:
Human beings possess unique physical characteristics that distinguish them from one another. According to ISO 27001, an effective Information Security Management System (ISMS) implements the principles of Confidentiality, Integrity, and Availability (CIA). To uphold these principles, identification and authentication are essential for accessing certain information. Identification involves verifying if an identity is registered in the system. Authentication, on the other hand, ensures the accuracy of the provided information and differentiates individuals from one another.
Authentication is categorized into three main types: