Electronic Theses and Dissertation

Penelitian ini membahas pengembangan sistem deteksi vegetasi mangrove dengan menggunakan arsitektur CenterNet yang dimodifikasi pada bagian backbone menjadi EfficientNet-B0. Studi dilakukan di kawasan mangrove Gampong Pande, Banda Aceh, yang terdiri atas tiga genus utama, yaitu Rhizophora, Avicennia, dan Nypa. Data penelitian dikumpulkan melalui citra UAV beresolusi tinggi dan dianotasi secara manual untuk menghasilkan dataset sebanyak 389 citra dengan 5.267 anotasi objek. Proses pelatihan dilakukan dengan konfigurasi 100 epoch, batch size 16, dan resolusi citra 640×640 piksel. Hasil evaluasi menunjukkan bahwa model yang dikembangkan mampu mencapai nilai mean Average Precision (mAP) sebesar 0,45, precision sebesar 0,78, dan recall sebesar 0,83. Temuan ini mengindikasikan bahwa modifikasi arsitektur CenterNet–EfficientNet-B0 dapat meningkatkan kinerja deteksi mangrove, meskipun keterbatasan dataset yang tidak seimbang masih memengaruhi hasil. Penelitian ini diharapkan dapat menjadi dasar pengembangan sistem deteksi mangrove berbasis UAV untuk mendukung konservasi dan pemantauan ekosistem pesisir secara lebih efektif.

This study discusses the development of a mangrove vegetation detection system using a modified CenterNet architecture, in which the backbone was replaced with EfficientNet-B0. The research was conducted in the mangrove area of Gampong Pande, Banda Aceh, which consists of three main genera: Rhizophora, Avicennia, and Nypa. Data were collected using high-resolution UAV imagery and manually annotated, resulting in a dataset of 389 images with 5,267 object annotations. The training process was carried out with 100 epochs, a batch size of 16, and an image resolution of 640×640 pixels. The evaluation results show that the developed model achieved a mean Average Precision (mAP) of 0.45, a precision of 0.78, and a recall of 0.83. These findings indicate that the modified CenterNet–EfficientNet-B0 architecture can improve mangrove detection performance, although dataset imbalance still affects the results. This research is expected to serve as a foundation for developing UAV-based mangrove detection systems to support more effective coastal ecosystem monitoring and conservation efforts.