IMPLEMENTASI METODE CONVOLUTIONAL NEURAL NETWORK UNTUK KLASIFIKASI RAS ANJING DENGAN ARSITEKTUR INCEPTIONV3

Abstract

 This research aims to develop an application capable of identifying dog breeds through graphic image detection using the Convolutional Neural Network (CNN) method. The implementation of artificial intelligence (AI) technology is increasingly widespread, including for pet breed recognition applications. The main challenge in conventional dog breed identification lies in the complexity of their physical variations, fur, and colors, requiring specific knowledge in the field of domestic animals for accurate results. By utilizing the CNN-based deep  learning algorithm in the development of this AI application, accurate recognition of unique patterns and  characteristics that distinguish each dog breed can be achieved through proper training with a large and comprehensive dataset. In this research, the model was trained on 2,758 images with an 80:20 dataset ratio, divided into 15 classes using the Inceptionv3 model architecture. During the training process, the model successfully achieved an F-1 Score above 85% for 13 out of 15 classes, with an overall final accuracy of 94.71%. The conclusion obtained is that the desktop-based dog breed recognition application works optimally in identifying dog breeds.   Keywords—Artificial Intelligence, Convolutional Neural Network, Dog Breed