<![CDATA[Indonesia is known as a country with very high biodiversity, one of which is reflected in the many bird species spread across various regions. The high public interest in birds presents challenges, in the form of increased bird trade practices, especially protected species. The main obstacle often faced by the public is the lack of information regarding the identification of protected bird species. Therefore, an application that can identify protected birds is needed. This study developed a web-based application using the Transfer learning approach in the classification of bird species in Indonesia. Five machine learning models were compared in this study: CNN, MobileNetV3Small, MobileNetV3Large, VGG16, and InceptionV4. The birds-525-species-image-classification dataset was used in the training process, then filtered to 54 bird species found in Indonesia. Model performance was evaluated using accuracy, precision, recall, and F1-score metrics. The developed web application uses React as the frontend and FastAPI as the backend and provides image upload and camera capture features to detect bird species directly. Test results show that MobileNetV3Large provides the most optimal performance and was selected for implementation in the system. In system testing through Black Box Testing, User Acceptance Test (UAT), System Usability Scale (SUS), and Net Promoter Score (NPS), it was shown that the application built had a good level of acceptance and ease of use]]>
Copyrights © 2025
