Deep Learning-Driven Facial Recognition for Secure, Keyless Vehicle Access and Personalization

Authors- Sivamanikandan B, Assistant Professor Nandhini K

Abstract--Traditional key- or PIN-code-based vehicle access systems are plagued with large security and convenience-related issues. The problem of concern here is the need for a secure, non- intrusive, and precise method of authenticating drivers and passengers. This paper presents a deep learning-based facial recognition system, and in particular, highlights the Convolutional Neural Network (CNN) mechanism bolstered by data augmentation methods to increase resistance against illumination variations, aging, and partial occlusion. The foremost goal is the security and convenience of the vehicle by enabling keyless access and automatic vehicle setting personalization. The system structure of the given system includes a data augmentation module, a training module, and real-time prediction and test modules, ensuring scalability and flexibility. The novelty in the work lies in combining CNN and Capsule Networks with advanced augmentation, real-time liveness detection, and auto vehicle setting adjustment based on detected individuals, thereby realizing a secure, efficient, and convenient approach for next-generation cars and eliminating the problems involved in real-world deployment.

DOI: /10.61463/ijset.vol.13.issue3.142