Change Detection of Floating Net Cages Quantities Utilizing Faster R-CNN

Abstract

The aquaculture method uses floating net cages are the most productive fish farming techniques. We utilize deep learning for change detection and monitoring of floating net cages quantities by detecting & counting the number of floating net cages plots on the same Region of Interest (RoI) in different years using google earth satellite imagery. The proposed methods apply Faster R-CNN for detection purposes and compare Faster R-CNN between using NASNet-A and inception-v2 as the feature extractor. Our experiments have been conducted on annotation images by cropping google earth images to demonstrate the effectiveness and efficiency of the proposed method. The results show that Faster R-CNN using NASNet-A achieves higher accuracy with longer training time. In addition, Faster R-CNN with inception-v2 network also provided promising results with lower training time.