DETEKSI JENIS RAS KUCING MENGGUNAKAN TRANSFER LEARNING MODEL EFFICIENTNET BERBASIS WEBSITE
Cat Breed Detection Using Transfer Learning with EfficientNet and a Web-Based Application
Kucing merupakan salah satu hewan peliharaan yang populer dengan berbagai ras yang memiliki karakteristik unik. Pengetahuan mengenai ras kucing penting bagi pemelihara untuk memahami kebutuhan spesifik seperti nutrisi, perilaku, dan potensi masalah kesehatan. Penelitian ini mengembangkan model klasifikasi ras kucing berbasis website menggunakan arsitektur EfficientNet-B0 dengan pendekatan transfer learning. Model dilatih pada dataset berisi 12 ras kucing populer dan diuji performanya melalui confussion matrix. Kombinasi EfficientNet B0 dengan optimizer Adam mendapatkan hasil paling optimal, dengan akurasi training set 92% , validation set 89%, dan 88% pada data uji. Selanjutnya, sistem web berbasis Flask yang mengintegrasikan model ini dievaluasi menggunakan metode Black-box Testing untuk memastikan keandalan fungsionalitas sistem. Hasil uji menunjukkan bahwa seluruh fitur berjalan sesuai spesifikasi dan model mampu memberikan prediksi ras dengan sesuai. Sistem ini diharapkan mempermudah pemelihara dan pecinta kucing dalam mengenali ras kucing di lapangan.
Cats are among the most popular pets, encompassing various breeds each with unique characteristics. Understanding cat breeds is essential for owners to meet specific needs such as nutrition, behaviour, and potential health issues. This study develops a web based cat breed classification model using the EfficientNet-B0 architecture with a transfer learning approach. The model was trained on a dataset of 12 common cat breeds and its performance evaluated via a confussion matrix. The combination of the EfficientNet-B0 and the Adam optimizer yielded the best result, achieving 92% accuracy on the training set, 89% on validation set, and 88% on the test set. Furthermore, the Flask based web application integrating this model was assessed through Black-box Testing to verify the reliability of system functionalies. The test results demonstrated that all features operated according to spesifications and that the model provided accurate breed predictions. This system is expected to assist cat owners and enthusiast in swiftly and accurately identifiying cat breeds in real world settings.