Willy, Achmat Fauzi (2021) Sistem Pengenalan Wajah Berbasis Deep Metric Learning dalam Skala Besar. Masters thesis, Institut Teknologi Sepuluh Nopember.
Text
07111750052003-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (6MB) | Request a copy |
Abstract
Penelitian ini menyajikan sistem pengawasan berbasis CCTV yang untuk pendeteksi dan pengenalan wajah pada uncontrolled environment. Peneltian ini memiliki dua kontribusi yaitu membuat objective lost function baru pada arsitektur deep metric learning yang kami sebut dengan coulomb loss. Kami mendapatkan accuracy tertinggi terhadap dataset LFW adalah sebesar 99.61% dengan data pelatihan terbatas. Kedua, diusulkan optimasi sistem verifikasi wajah dengan multiple face detection menggunakan kalman dan hungarian algorithm. Dari hasil percobaan,dengan algoritma kami 87.832 wajah yang seharusnya dikenali berkurang menjadi hanya 204 wajah, dan sistem berjalan secara real-time.
====================================================================================
This research presents a CCTV-based surveillance system for face detection and recognition in uncontrolled environment. This research has two contributions, creating a new objective lost function in a deep metric learning architecture which we call coulomb loss. We get the highest accuracy against the LFW dataset of 99.61%. Second, it is proposed to optimize the face verification system with multiple face detection using Kalman and the Hungarian algorithm. Based on the result, using our algorithm 87.832 face that must be recognized is reduced to only 204 faces, and run at the realtime scenario.
Item Type: | Thesis (Masters) |
---|---|
Uncontrolled Keywords: | face recognition, deep metric learning, coulomb loss, smart city |
Subjects: | T Technology > T Technology (General) |
Divisions: | Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Electrical Engineering > 20101-(S2) Master Thesis |
Depositing User: | Willy Achmat Fauzi |
Date Deposited: | 26 Feb 2021 04:07 |
Last Modified: | 26 Feb 2021 04:07 |
URI: | http://repository.its.ac.id/id/eprint/82897 |
Actions (login required)
View Item |