Generating Information of URL Based on Web Scraping Using YOLOv3 Face Recognition Technology

Lulud Annisa Ainun Mahmuddah; Suryo Adhi Wibowo; Gelar Budiman

doi:10.25124/ijait.v5i02.3910

Lulud Annisa Ainun Mahmuddah The University Center of Excellence for Advanced Intelligent Communications, Telkom University, Indonesia; Image Processing and Vision Laboratory, Telkom University, Indonesia; School of Electrical Engineering, Telkom University, Indonesia http://orcid.org/0000-0001-8710-5523
Suryo Adhi Wibowo The University Center of Excellence for Advanced Intelligent Communications, Telkom University, Indonesia; Image Processing and Vision Laboratory, Telkom University, Indonesia; School of Electrical Engineering, Telkom University, Indonesia
Gelar Budiman School of Electrical Engineering, Telkom University, Indonesia

DOI: https://doi.org/10.25124/ijait.v5i02.3910

Abstract

Artificial Intelligence (AI) is a system developed to learn and apply human intelligence. Some technologies produced from the development of Al are face recognition and web scraping. Face recognition is used for identifying or verifying the identity of an individual using their face. The result of a face recognition process can be used to collect information on the internet with a web scraping technique. This paper proposes a face recognition model and web scraping system using You Only Look Once (YOLO) object detection method and Request library written in Python. The face recognition model performed fine-tuning in two hyperparameters, which are learning rate and step training. The proposed model for face recognition is using custom datasets that contain 8000 images divided into 5 classes and evaluated using the Mean Average Precision (mAP) performance parameter, while the web scraping system is evaluated using the precision rate parameter. From the test results, the best configuration was obtained at a learning rate of 0.0001 and step training of 10K. The highest mAP that is achieved is 0.90 with a recall and precision value of 0.75 for each, while the average precision rate is 0.87. The results of this paper are expected to contribute to the development of biometric security technology.

Downloads

Download data is not yet available.