Transfer Learning and Fine-tuning Effect Analysis on Classification of Cat Breeds using a Convolutional Neural Network

Cats are the most popular pets because they are adorable, entertaining, and can be used as companions when lonely. Its attractive appearance is one of the most popular animals in the world. However, various cat breeds in the world have different physical characteristics. To find out the type of cat breed, you can do a Deoxyribonucleic Acid (DNA) test. Unfortunately, DNA testing is costly and takes a long time. With Artificial Intelligence technology, a system is needed to classify cat breeds, namely the Convolutional Neural Network (CNN). This research aims to analyze the effect of transfer learning and fine-tuning on the classification of cat breeds using CNN. This research used a dataset from Oxford-IIIT Pet, which consisted of 12 classes, namely Abyssinian, Bengal Birman, Bombay, British Shorthair, Egyptian Mau, Maine Coon, Persian, Ragdoll, Russian Blue, Siamese, and Sphynx with a total of 2,400 images. This research uses parameters with an architecture using Xception and RMSprop optimization with a learning rate 0.0001. This research has two scenarios: CNN with transfer learning and CNN with transfer learning and fine-tuning. This research achieved 92.5% accuracy on CNN with transfer learning and fine-tuning.