Coral reefs are marine ecosystems that are highly vulnerable to damage and require regular monitoring of their health conditions. However, the manual classification process of coral reef health tends to be time-consuming. Therefore, this research aims to develop an application that implements a transfer learning model for classifying coral reef health based on digital images. This study utilizes three pretrained model architectures: DenseNet121, MobileNetV2, and EfficientNet-B0. Each model is trained and evaluated to measure its performance in classifying coral reef images. The best-performing model, DenseNet121, is then integrated into a mobile application for real-time classification. The evaluation results show that DenseNet121 achieved the highest accuracy compared to MobileNetV2 and EfficientNet-B0. The training data accuracy of DenseNet121 reached 98.80%, and the testing data accuracy was 98.25%.
Bahasa Indonesia
Terumbu karang merupakan ekosistem laut yang rentan terhadap kerusakan dan membutuhkan pemantauan rutin terhadap kondisi kesehatannya. Namun, proses klasifikasi kesehatan terumbu karang secara manual cenderung memakan waktu. Oleh karena itu, penelitian ini dilakukan untuk menghasilkan sebuah aplikasi yang mengimplementasikan model transfer learning dalam klasifikasi kesehatan terumbu karang berbasis citra digital. Penelitian ini menggunakan tiga arsitektur model pretrained, yaitu DenseNet121, MobileNetV2, dan EfficientNet-B0. Setiap model dilatih dan dievaluasi untuk mengukur kinerjanya dalam mengklasifikasikan citra terumbu karang. Model dengan kinerja terbaik yaitu DenseNet121, kemudian diintegrasikan ke dalam aplikasi mobile untuk klasifikasi secara langsung. Hasil evaluasi menunjukkan bahwa model DenseNet121 memperoleh akurasi terbaik dibandingkan MobileNetV2 dan EfficientNet-B0. Akurasi data training DenseNet121 sebesar 98,80% dan data testing sebesar 98,25%.
Collaboration network of the authors
17 authors ·
28 co-authorship links
Hover or tap a name to highlight its co-authors. Double-click to open the author's page.
Universitas Halu Oleo
Affiliation unknown
Different organisations
Same organisation
Affiliation unknown
Thicker line = more joint articles · bigger circle = more articles
Affiliations come from OpenAlex; journal OAI-PMH metadata carries none. Authors without an OpenAlex ID are matched by name, which can merge or split people.