Abstract
Accurate monitoring of polyp attachment density is critical for the efficient culture of the edible jellyfish Rhopilema esculentum , yet quantitative density–growth guidelines remain limited. This study coupled image-based deep learning with conventional morphometry to characterise density-dependent growth of polyps. Polyp calyx diameter was measured manually to establish a density–size relationship. In parallel, standardised photography combined with a U-Net segmentation model was used to obtain individual polyp counts and projected areas. Regression analyses were conducted to derive functional relationships between attachment density and polyp size metrics, and processing times were compared between manual and automated approaches. Manual measurements showed that calyx diameter followed a power-law decay with density ( Calyx diameter = 1.5752Density -0.281 , R 2 = 0.9614). Automated image analysis yielded an exponential density–polyp area model ( Polyp area = 4.3888e -0.202 Density , R 2 = 0.9909). Both models revealed a strong inverse relationship: as attachment density increased, average polyp size and relative growth efficiency declined, whereas size variability increased. The U-Net-based segmentation approach achieved processing times of under one second per image, dramatically faster than manual measurements. These results demonstrate that AI-driven image segmentation provides accurate, high-throughput estimation of polyp size and robust quantitative density–growth relationships. The approach offers a practical and efficient tool for precision monitoring and optimisation of nursery conditions in R. esculentum aquaculture.