DESIGN OF FACE RECOGNITION ALGORITHM FOR LESSCONTACT ATTENDANCE MANAGEMENT SYSTEM.

The COVID-19 pandemic has caused changes in human daily activities such as wearing masks when leaving the house, avoiding crowds, working from home and much more. But offices in the essential sector still have to open offices and workers must work from the office and covid-19 can be transmitted t...

Full description

Saved in:
Bibliographic Details
Main Author: Yapin, Ferdinand
Format: Final Project
Language:Indonesia
Online Access:https://digilib.itb.ac.id/gdl/view/58284
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Institut Teknologi Bandung
Language: Indonesia
Description
Summary:The COVID-19 pandemic has caused changes in human daily activities such as wearing masks when leaving the house, avoiding crowds, working from home and much more. But offices in the essential sector still have to open offices and workers must work from the office and covid-19 can be transmitted through physical contact with carriers or physical contact with virus residues left on objects. This raises various problems, one of which is the recap of employee attendance data. The recap of employee attendance data must be done without need of contact in order to avoid the transmission of the covid-19 virus. Less-Contact Presence System designed using an embedded system in the form of a Single Board Computer Nvidia Jetson Nano. Face recognition algorithms are usually implemented on computers that have high computing power, so implementing face recognition algorithms in embedded systems is a big challenge. It takes a face recognition algorithm that is lightweight and can run with limited resources. The design of this face recognition algorithm uses a siamese network and consists of 3 parts, namely Face Detection, Feature Extraction, and Classifier. Face detection uses an efficient pretrained Convolutional Neural Network (CNN) model, Faceboxes. Feature Extraction using mobilefacenet which is a lightweight CNN model and can be run on smartphones. From the tests carried out, the designed face recognition algorithm obtains an accuracy of 96.96% with an inference/performance speed of 1.5 s.