Identifikasi Tanaman Air terhadap Kebutuhan Cahaya, CO2, dan Penempatan dalam Aquascape Menggunakan YOLO

Abstract

Designing a balanced and aesthetically pleasing aquascape is often hindered by difficulties in identifying and placing aquatic plants according to their specific light, CO2, and height requirements, with manual processes being time-consuming and demanding high expertise. This research aims to develop an Android mobile application for real-time identification, classification, and placement recommendation of aquatic plants using the efficient YOLOv11 (You Only Look Once version 11) nano algorithm for mobile devices. The methodology involved collecting 4,095 aquatic plant images categorized into seven classes, data preprocessing including annotation and augmentation, training the YOLOv11n model for 50 epochs, and developing the application with Kotlin and TensorFlow Lite. Training results demonstrated excellent model performance, achieving a precision of 0.991, recall of 0.996, mAP@0.5 of 0.995, and mAP@0.5:0.95 of 0.967. The developed application has a precision value of 0.857 for the real-time identification feature and 0.893 for identifying images from the gallery. The developed application successfully identifies plants, presents their requirements, and recommends placement (foreground, middleground, background), thereby assisting aquascapers in creating optimal and visually appealing aquascapes.