Cat Breed Classification Based on Ear and Facial Features Using K-Nearest Neighbors

Abstract

Identifying cat breeds is often challenging, mainly due to the similarity of physical features between breeds and the high frequency of crossbreeding. This study aims to develop a cat breed classification system based on facial features, especially ear shape and facial structure, using the K-Nearest Neighbors (K-NN) algorithm. Five cat breeds—Bengal, British Shorthair, Mainecoon, Persian, and Sphynx—were used as test objects with 50 test data. The evaluation results show that the K value dramatically affects the system's accuracy, with the highest accuracy of 90% achieved at K values = 4, 5, and 7. The Bengal breed showed the highest classification accuracy of 100%, while the Sphynx breed performed the lowest in several scenarios. These findings confirm that facial features are relevant and effective parameters in cat breed identification and that the K-NN method can be a lightweight yet accurate classification solution. This study contributes to developing an image recognition system based on specific visual features for pet classification.