Diouf, A.; Bereziat, L.; Nodier, D.; Amaral, M.; Haliyo, S.; MANNIOUI, A.

Accurate quantification and continuous monitoring of Brachionus plicatilis L. rotifer cultures are essential for aquaculture and aquatic animal research laboratories. Manual counting methods are labor-intensive, error-prone, and inefficient for large-scale operations, necessitating automated solutions. This study presents the Rotiferometer, an automated and cost-effective system that integrates mechanical design, deep learning, and automation for precise rotifer detection, classification and counting. Using a YOLOv8 model, the system achieves a mean average precision ([email protected]) of 94.7% in distinguishing gravid and non-gravid rotifers. It proceeds by scanning a 1 mL Sedgewick Rafter slide under 3 minutes, ensuring rapid and accurate enumeration. A strong correlation was observed between manual and Rotiferometer counts, (with R2 values of 0.9729 and 0.9868 for gravid and non-gravid rotifers, respectively), confirming the system s accuracy. Additionally, the analysis of operator variability using the Rotiferometer delivered consistent results regardless of the user, minimizing the need for specialized expertise. Finally, a 45-day monitoring experiment with the Rotiferometer effectively tracked rotifer population changes, identifying key phases of growth, decline, and recovery. These results highlight the device s potential to enhance rotifer culture management by providing real-time, reliable, and automated monitoring, thereby optimizing aquaculture productivity and research efficiency.