Disease Segmentation in Water Hyacinth (Eichhornia Crassipes) Images Using YOLOv8

Water hyacinth is an aquatic plant from the Amazon River. Water hyacinth is a serious weed in freshwater environments in rivers, lakes and reservoirs in warm and tropical climates worldwide. Water hyacinth can also reduce water quality and quantity by reducing dissolved oxygen. Although considered a weed, the roots of Eichhornia crassipes can absorb contaminated water, including heavy metals such as lead and mercury and some organic composites that are confirmed carcinogenic. Therefore, Eichhornia crassipes can be cultivated for wastewater treatment. One of the problems encountered in managing this aquatic plant is identifying and monitoring diseases that can affect plant growth and health. Certain diseases can interfere with plant growth, therefore a way is needed to identify water hyacinth diseases so that plant management becomes efficient using the right technology. this study aims to create a disease segmentation system in water hyacinth using YOLOv8 and analyze the performance of YOLOv8 in identifying diseases in water hyacinth leaves, from the results of testing water hyacinth leaf disease segmentation. It can be concluded that segmentation of water hyacinth leaves using YOLOv8 with maximum results according to the disease of each water hyacinth leaf, starting from a healthy leaf (daun sehat) has an mAP value of 0.995. Then leaves with spot disease (daun berpenyakit bercak) have a water hyacinth leaf segmentation result of 0.977. Rust-diseased leaves (daun berpenyakit karat) have a value of 0.928, yellow-diseased (daun berpenyakit kuning) leaves have a value of 0.995 and the value results for all classes obtained are 0.974. In doing segmentation it takes a relatively fast time, namely Speed: preprocess 5.1ms, inference 14.1ms, loss 0.0ms, postprocess 3.7ms per image.