<![CDATA[Traditional image classification research has focused on single-input and multiclass approaches. However, these approaches often fail to capture the complexity and diversity of real-world image data. To address the complexity and more diverse variation in data, as well as to improve the classification accuracy of various categories, a multi-input image approach is utilized. With a multi-input multi-class approach, a Transfer Learning model based on VGG16 is trained to identify objects from various perspectives and classify them into one of many predefined classes. The VGG16 architecture in the multi-input and multi-class classification of Toraja Buffalo breeds demonstrates excellent results with an average accuracy of 93.33%. The "Kerbau Lotong Boko" and "Kerbau Bonga Ulu" classes achieved 100% accuracy, while other classes showed high precision, recall, and F1 scores. Despite fluctuations in accuracy and loss during training, the model successfully achieved good convergence and generalization. This research is significant in the field of image classification by introducing a multi-input method capable of capturing richer and more diverse information from complex objects such as Toraja buffalo. It demonstrates that CNN architectures like VGG16 can be adapted to handle more complex classification tasks using a multi-input approach.]]>
Copyrights © 2024
