Identification of Different Types of Buffalo Using YOLO Algorithm Based on Computer Vision

Abstract

The purpose of this research is to create a system that can distinguish and correctly label different types of black and white buffaloes, such as Bonga, Saleko, Lotongboko, and buffaloes that are not black and white or other than their species. This research uses classification methods in pattern recognition, training a model to distinguish the different characteristics of each type of buffalo. This model was then used to classify the buffalo species in the images using the YOLO (You Only Look Once) algorithm. The training involved 40 learning iterations. The dataset consisted of 497 buffalo images, each measuring 640 x 640 pixels, which were divided into 442 images for training and 55 images for validation. By the 40th epoch, the model had achieved an overall mAP50 value of 0.959 and mAP50-95 of 0.825. The research shows that the buffalo species detection and classification system achieved optimal performance, demonstrating the effectiveness of the YOLO algorithm in computer vision, achieving 98% accuracy in the identification and labeling of different buffalo species using image analysis and pattern recognition techniques, while the increasing mAP value indicates that the model is still learning and improving its performance with each additional epoch. This study demonstrates the very high accuracy and potential convergence of the model to optimal performance.

Access to Document

Other files and links

• Link to publication in Scopus

Fingerprint