Instance Segmentation-Based Shrimp Weight Predictor using YOLOv8 and Bayesian Regression Model

Shrimp Farmers traditionally measure shrimp’s size by picking them from the pond using a net, putting them into scales, drying them, and counting the average body weight. This method is time-consuming. Additionally, continuous and high-frequency sampling, which is essential for accurate shrimp growth prediction and efficient shrimp farming, is challenging with traditional methods. In this paper, We developed a method of shrimp weight sampling using computer vision and machine learning model. The proposed system achieves an 11.74% MAPE score for individual weight prediction and a 6.59% MAPE score for ABW prediction with a 96% detection rate.