<![CDATA[Cattle breed identification presents a challenge in modern livestock management, particularly for breeds exhibiting high morphological similarities. This study aims to optimize cattle breed image classification performance using the ResNet-50 Convolutional Neural Network (CNN) architecture. The research methodology begins with preprocessing, which includes image resizing to 224×224 pixels, normalization, and the application of Data Augmentation to enhance training data diversity. The model was trained using the Adam optimizer with specific hyperparameter setting scenarios. The primary focus of this research is to evaluate the effectiveness of the Transfer Learning method compared to training from scratch. The dataset consists of 1,251 morphological images distributed across 5 breed classes (Ayrshire, Brown Swiss, Holstein Friesian, Jersey, and Red Dane). Experimental results demonstrate a significant performance disparity. The model trained from scratch achieved a maximum accuracy of only 81%, whereas the Transfer Learning-based model attained an accuracy of 95%. This 14% accuracy improvement demonstrates that utilizing pre-trained weights combined with Data Augmentation is highly effective in recognizing the visual characteristics of cattle breeds with high precision.]]>
Copyrights © 2026
