OPeraTE.AI: Optimized Personnel and Inmate Tracking Efficiency Through Facial Recognition using Siamese Neural Network

Abstract

The Philippine National Police (PNP) in Clarin, Bohol, faces significant challenges in managing personnel and inmate information due to disorganized files, manual processes, and a lack of a centralized system. This inefficiency leads to data redundancy, security risks, and time-consuming retrieval of information. To address issues, the development of a web-based, centralized system, Optimized Personnel and Inmate Tracking Efficiency (OPeraTE), designed to streamline personnel and inmate profiling and improve data security was proposed. The system uses a Siamese Neural Network (SNN) with Triplet Loss to Implement accurate facial recognition, allowing the identification and retrieval of profiles using text and images using K-Fold in training to obtain the hyperparameter values needed to make the model accurate. The SNN architecture consists of three convolutional layers that process input images (anchor, positive, and negative) to generate embeddings, which are then compared to calculate similarity via Squared Euclidean distance. This enables identification with minimal data. OPeraTE includes functionalities like profile creation, editing, search through text and face, and archiving. Performance metrics, including accuracy (78.20%), precision (73.80%), recall (46.88%), F1 Score (57.34%), False Acceptance Rate (FAR) (7.56%), and False Rejection Rate (53.12%), demonstrate the systems efficiency in identifying personnel and inmates. The system was deployed in the PNP-Clarin station, providing a more efficient and secure method of managing personnel and inmate profiles. The results confirm that OPeraTE significantly enhances data management, reduces errors, and contributes to improved law enforcement operations and public safety.