Feeding Behavior Classification of Nile Tilapia (Oreochromis niloticus) using Convolutional Neural Network

Abstract

Feeding management in aquaculture is very important and has a big impact to the farmers and environment. A good feeding management optimizes growth and feeding efficiency. It also decreases the amount of excess nutrients entering the environment. Several existing systems aimed to provide an efficient and effective feeding management using image processing and deep learning. This study developed a model that will be part of a fish feeding management system that used fish feeding behavior to know the feeding state of the fish. The methodology used was image generation, image processing, classification and testing. Convolutional Neural Network (CNN) was utilized to classify fish feeding behavior into two states; to feed or not to feed. The CNN model was tested for accuracy, precision, recall, and specificity and the results were 96.4%, 97.87%, 94.867% and 97.93%, respectively. The result of the study will be used and integrated to a feeding management system.