Klasifikasi Buah Mangrove Berdasarkan Spesies Menggunakan Metode SSD-MobileNet Berbasis Mobile secara Real-Time

Abstract

Mangrove are plants that live in intertidal zones and provide various ecological and economic benefits, such as preventing coastal erosion and providing habitats for coastal fauna. Classifying mangrove species based on their fruits is quite challenging due to the similarity in shape among species. This study aims to develop a system that can recognize mangrove fruits in real-time by implementing the Single Shot Detector MobileNet (SSD-MobileNet) method on an android platform to assist farmers and the community in accurately classifying mangrove fruits. The method used in this research involves capturing and pre-processing images of mangrove fruits with an input size of 640 × 640 pixels after resizing. The dataset consists of 1440 training data, 270 validation data, and 90 testing data. This study aims to classify three mangrove species: Rhizophora mucronata, Rhizophora stylosa, and Rhizophora apiculata. The system implementation is carried out on android device to ensure easy access and use in the field, especially for farmers and the community who need practical tools for mangrove species classification. The outcome indicates that the system attained an accuracy of 95,56% when tested using a smartphone camera in real-time. This accuracy rate indicates that the SSD-MobileNet method is effective in classifying mangrove fruits.