Improving accuracy and efficiency in seagrass detection using state-of-the-art AI techniques

Seagrasses provide a wide range of ecosystem services in coastal marine environments. Despite their ecological and economic importance, these species are declining because human impact. This decline has driven the need for monitoring mapping to estimate overall health dynamics seagrasses environments, often based on underwater images. However, seagrass detection from digital images is not trivial task; it requires taxonomic expertise time-consuming expensive. Recently automatic approaches deep learning have revolutionised object performance many computer vision applications, there been interest applying this automated imagery. Deep learning–based techniques reduce hardcore feature extraction by domain experts which required machine learning-based techniques. study presents YOLOv5-based one-stage detector an EfficientDetD7–based two-stage detecting seagrass, case, Halophila ovalis, one most widely distributed species. The EfficientDet-D7–based achieves highest mAP 0.484 ECUHO-2 dataset 0.354 ECUHO-1 dataset, about 7% 5% better than state-of-the-art ovalis those datasets, respectively. proposed average inference time 0.077 s 0.043 respectively much lower approach same datasets.