EVALUASI PENGARUH FINE-TUNING DAN DATA AUGMENTATION TERHADAP KINERJA MOBILENETV2 DAN RESNET50 PADA KLASIFIKASI RAS KUCING MENGGUNAKAN TRANSFER LEARNING: EVALUATION OF THE EFFECTS OF FINE-TUNING AND DATA AUGMENTATION ON MOBILENETV2 AND RESNET50 PERFORMANCE IN CAT BREED CLASSIFICATION USING TRANSFER LEARNING

Imam Munzagi (Universitas Muria Kudus)
Yudie Irawan (Universitas Muria Kudus)
Fajar Nugraha (Universitas Muria Kudus)

RABIT: Jurnal Teknologi dan Sistem Informasi Univrab
Vol 11 No 2 (2026): Juli

Manual identification of cat breeds often faces challenges due to the high similarity of visual features among breeds, potentially leading to errors in medical treatment and nutritional management. This study aims to compare the performance of the MobileNetV2 and ResNet50 architectures in classifying 13 cat breeds using a Transfer Learning approach. The CRISP-DM methodology was implemented through four experimental scenarios to evaluate the effects of Data Augmentation and Fine-Tuning, both individually and in combination. The results indicate that the combination of Data Augmentation and Fine-Tuning achieved the best performance. MobileNetV2 consistently outperformed ResNet50 across all scenarios, achieving the highest accuracy of 90,00% and an F1-Score of 90,02%, while ResNet50 achieved a maximum accuracy of only 39,69%. The 50,38% performance gap demonstrates that MobileNetV2 is more adaptive in extracting visual object features. The best-performing model was successfully implemented into an application prototype that is functional and user-friendly.