Tilapia Fish Weight Estimation in Freshwater Using Deep Learning and Regression Methods

This study presents a 2D computer vision approach to estimate Tilapia weight in freshwater using a multi camera for video & picture capturing. The method relies on a custom Tilapia image dataset, including various fish growth stages, and involves two steps: detecting Tilapia with a Mask R-CNN model and estimating weight. Depth is calculated to convert pixel dimensions to centimeters, followed by weight prediction using regression models. Linear regression, random forest, and support vector regression were tested for accuracy. The approach eliminates the need for expensive stereo cameras, providing an efficient, non-intrusive solution.