Komparasi Performa Klasifier CNN Untuk Re-identifikasi Kendaraan Roda Empat

Kurniawan, Agung (2021) Komparasi Performa Klasifier CNN Untuk Re-identifikasi Kendaraan Roda Empat. Undergraduate thesis, Institut Teknologi Sepuluh Nopember.

Text 07211640000048-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2023. Download (12MB) | Request a copy

Abstract

Re-identifikasi adalah sebuah proses pencarian ulang sebuah informasi yang masuk untuk mencocokan dengan data yang sudah
ada. Proses re-identifikasi dapat dilakukan dengan menggunakan deep learning, salah satunya menerapkan arsitektur Convolutional Neural Network (CNN) dengan berbagai macam model. Penerapan re-identifikasi kendaraan roda empat pada kamera closed circuit television CCTV di Indonesia juga masih tergolong baru. Jika sistem ini ingin dikembangkan, maka harus menentukan model CNN manakah yang paling efektif untuk diterapkan pada re-identifikasi kendaraan roda empat. Sehingga dilakukan penelitian komparasi antar metode CNN berdasarkan tingkat performansi pada kendaraan roda empat. Penelitian ini bertujuan untuk menemukan metode CNN yang efektif dalam melakukan re-identifikasi kendaraan roda empat, dengan membandingkan beberapa metode yang umum digunakan untuk melakukan re-identifikasi. Pada penelitian ini kami mengujikan beberapa model CNN, yaitu ResNet50,ResNet50+Circle Loss, ResNet50+CPB, InceptionResNetv2,DenseNet121, EficientNet, dan NasNetLarge, dari beberapa model tersebut kami mendapatkan nilai Rank1 sebesar 35%, Rank5 37%,Rank10 49%,dan mAP sebesar 16%, pada model ResNet50+CPB dan merupakan nilai tertinggi dari model lain pada pengujian.
================================================================================================
Re-identification is a process of re-finding an incom- ing information to match the existing data. The re-identification process can be done using deep learning, one of which is applying the Convolutional Neural Network (CNN) architecture with various models. The application of re-identification of four-wheeled vehicles on CCTV cameras in Indonesia is still relatively new. If this system is to be developed, it must determine which CNN model is the most effective to be applied to the re- identification of four wheeled vehicles. So that a comparative study was conducted between CNN methods based on the level of performance on four-wheeled vehicles. This study aims to find an effective CNN method for re-identifying four-wheeled vehicles, by comparing several methods commonly used to re-identify. In this study, we tested several CNN models, namely ResNet50, ResNet50+Circle Loss, ResNet50+CPB, InceptionRes- Netv2, DenseNet121, EfficientNet, and NasNetLarge, from these models we got Rank1 values off 35%, Rank5 37%, Rank10 49%, and mAP of 16%, on the ResNet50 + CPB model and is the highest value of the other models in the test.

Item Type:	Thesis (Undergraduate)
Uncontrolled Keywords:	Transportasi, CNN, Re-identifikasi, Kriminalitas, Deep Learning, Transportation, Re-identification, Criminality
Subjects:	R Medicine > R Medicine (General) > R858 Deep Learning
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User:	Agung Kurniawan
Date Deposited:	01 Sep 2021 03:46
Last Modified:	01 Sep 2021 03:46
URI:	http://repository.its.ac.id/id/eprint/91106

Actions (login required)

View Item