Biometric analysis in fish farms, such as weighing and measuring Nile tilapia, is a manual and stressful procedure for the animals and incurs significant costs, leading to growth delays and stock losses. This study investigates the application of Convolutional Neural Networks (CNNs) to estimate tilapia weight from images of its phenotype, with the aim of automating and optimizing this process. Using an existing image dataset, several CNN models were trained and evaluated. The results are promising, with some configurations achieving discrepancies of only 45–50 grams between predicted and actual weight, and a coefficient of determination (????2) of 0.97. It is concluded that it is feasible to develop an efficient and accurate system to predict the weight of Nile tilapia using CNNs, with potential to increase productivity and reduce costs in fish farming.