Analisis Perbandingan VGG-16 dan ResNet50 untuk Klasifikasi Multilabel Gambar Kerbau Toraja: Pendekatan Deep Learning

Tri Anita Resky Ramadhani
Abdul Rachman Manga

Jurnal Teknik
Vol 22 No 2 (2024): Jurnal Teknik

This study aims to compare the performance of two Convolutional Neural Networks (CNN) models, namely VGG-16 and ResNet50, in the task of multilabel classification of buffalo images. The Dataset used consists of 2009 buffalo images labeled with five categories: human, motorcycle, truck, wild animal, and buffalo. The CNN models were trained using 30 epochs and evaluated using loss, accuracy, Precision, recall, and f1-score metrics. The experimental results show that VGG-16 consistently outperforms ResNet50 by achieving the highest accuracy of 0.95 in the training set and 0.94 in the validation set, and f1-score of 0.94 in the training set and 0.92 in the validation set. These findings indicate that a deeper and more structured CNN architecture, such as VGG-16, provides better results in classifying buffalo images with complex label variations.