Revolutionizing Aquatic Epidemiology: A Scalable Deep Learning Framework for Disease Detection and Enhancing Environmental Resilience

The sustainability of the lucrative fishing industry is contingent upon the health of fish stocks in the aquatic environment, which is also essential for maintaining ecological balance. The productivity and quality of fish are adversely affected by the absence of contemporary, technology-based, high- speed, and precise disease detection methods. In this context, an enhanced automatic fish disease detection method is proposed, utilizing a hybrid of ensemble learning and transfer learning techniques. Sophisticated pre-trained neural networks, including EfficientNetV2B0, EfficientNetV2S, and ResNet50, are employed for feature extraction and feature fusion in the proposed model. Experimental results demonstrate that the individual instances of EfficientNetV2B0 and EfficientNetV2S achieve an accuracy of 99.57%, while the fused model (EfficientNetB0+ResNet50) achieves 99.71% accuracy. The proposed ensemble model exhibits a maximum accuracy of 99.83%, representing a considerable advance over the current state of the art. Our proposed stan- dalone model has demonstrated superior performance compared to most modern fusion-based architectures and even ensemble frameworks. Through the efficient implementation of feature integration and transfer learning, the detection methods speci- ficity and sensitivity are substantially enhanced. The systems straightforward API simplifies deployment and sharing in real life. This API lets users upload fish photos, which the sys- tem processes and accurately diagnoses by disease type. This connection increases usability and makes the system a useful tool for intelligent, real-time farmed fish disease monitoring. By implementing scalable API-driven access and developing data- driven, intelligent methods for fish disease detection, this research has the potential to revolutionise aquatic health management and introduce a new era of technology-based aquaculture.