Implementasi Metode Convolutional Neural Network untuk Klasifikasi Ras Anjing

Fauzan, Muhammad (2023) Implementasi Metode Convolutional Neural Network untuk Klasifikasi Ras Anjing. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 05111640000075-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 July 2025. Download (1MB) | Request a copy

Abstract

Anjing adalah salah satu hewan peliharaan yang banyak disukai oleh banyak orang. Namun dengan banyaknya jumlah anjing menyebabkan beberapa masalah, seperti pengendalian populasi, vaksinasi, dan salah satu pembawa penyakit rabies. Saat ini, terdapat lebih dari 150 ras anjing. Setiap ras anjing memiliki perlakuan dan tingkat imunitas yang berbeda. Oleh karena itu, diperlukan suatu sistem yang dapat mengenali ras anjing dengan akurat, agar perlakuan terhadap anjing tersebut tepat. Pengenalan ras anjing dapat dilakukan dengan metode Convolutional Neural Network (CNN). Untuk mendapat hasil yang maksimal, dibutuhkan dukungan dari arsitektur dan hyperparameter nya agar mendapat hasil yang optimal. Arsitektur CNN yang digunakan adalah InceptionV3, ResNet50V2, dan MobileNetV2 dengan bantuan parameter transfer learning ImageNet, dan dropout. Model ini akan menggunakan data citra sebanyak 13,513 yang mencakup 15 ras anjing. Hasil dari pengenalan ras anjing ini memiliki akurasi paling baik dengan menggunakan arsitektur MobileNetV2, optimizer Adam dan nilai dropout sebesar 0.3, dengan nilai akurasi 98%, presisi sebesar 98.18%, recall sebesar 98% dan f1score sebesar 98%.
=====================================================================================================================================
Dogs are one of the pets that many people like. However, the large number of dogs causes several problems, such as population control, vaccination, and one of the carriers of rabies. Currently, there are more than 150 breeds of dogs. Each dog breed has a different treatment and level of immunity. Therefore, we need a system that can accurately identify dog breeds, so that the treatment of these dogs is appropriate. Dog breed recognition can be done using the Convolutional Neural Network (CNN) method. To get maximum results, support from the architecture and hyperparameters is needed to get optimal results. The CNN architectures used are InceptionV3, ResNet50V2, and MobileNetV2 with the help of ImageNet transfer learning parameters, and dropouts. This model will use 13,513 image data which includes 15 dog breeds. The result of dog breed classification has the best accuracy using the MobileNetV2 architecture, Adam optimizer and dropout value of 0.3, with an accuracy of 98%, precision of 98.18%, recall of 98% and f1score of 98%.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Convolutional Neural Network, Klasifikasi Citra, Pengenalan Ras Anjing, Transfer Learning, Classification, Dog Breed Classification
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. R Medicine > R Medicine (General) > R858 Deep Learning
Divisions:	Faculty of Information Technology > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Muhammad Fauzan
Date Deposited:	12 Feb 2023 12:15
Last Modified:	12 Feb 2023 12:15
URI:	http://repository.its.ac.id/id/eprint/96881

Actions (login required)

View Item