Aqua-Farm fish behavior analysis using machine learning

Aquaculture production has been shown to represent approximately 46% of the total fish production in 2018 from FAO. This percentage shows how aquaculture contributes to global fish production, and how the demand is increasing for fish and seafood products. Understanding Fish behavior is crucial for better fish production. Fishes showing abnormal behavior can serve as early warning signs of health issues, therefore monitoring behaviors such as feeding is important to ensure the growth rate of fishes. Machine learning provides a way to be used for classifying behaviors, then the results will be passed to the manager to notify about the status of the fish farm. In this paper we proposed a system which used RNN architectures(RNN-LSTM-BILSTM-GRU) and LRCNN for classifying the behaviors based on trajectories, and the architecture that achieved the biggest average accuracy and lowest validation loss with 100% and 0.00275 respectively using LRCNN was tested on Self collected fish Dataset.