Faster R-CNN based Fish Detector for Smart Aquaculture System

Abstract

One of the potential applications of computer vision and deep learning is object detection. Faster R-CNN was utilized in this work to create a fish detector that locates occurrences of fish in a frame. The performance of the developed model was evaluated using accuracy, root mean square error (RMSE) and intersection over union (IoU). After training and validation, the developed model achieved a mini batch accuracy equal to 99.95 percent with RPN mini batch accuracy equal to 100 percent. The system has a mini batch RMSE equal to 0.12 with RPN mini batch RMSE equal to 0.28. The computed mean IoU is equal to 0.7816.