Automated Monitoring System for Healthier Aquaculture Farming

Maintaining a healthier aquaculture system is difficult and without regular monitoring, significant losses may occur. The condition of the pond is a key concern and if the sites are left uncleaned, the number of dead fish will grow over time. Manually identifying these dead fish and consistently keeping aquaculture facilities in good condition is difficult, hence automation is the solution. This research offers a reliable model that can be applied to a UAV surveillance system and has been trained on thousands of images to recognize dead fishes floating over the surface, indicating an imbalance in the pond. By alerting the farmer of the presence of dead fish, immediate steps could be made to mitigate its impacts. We used the YoloV5 architectures to effectively detect dead-fish objects in real time and built the model to an accuracy of 88% considering various parameters like low light conditions and cropped versions of the fish image.