EFFICIENT UNDERWATER IMAGE CLASSIFICATION WITH ENHANCED YOLOV5 AND BEARDED DRAGON OPTIMIZATION (BeardYOLO)

The proposed work focuses on the enhancement of the EY5 model through the integration of BDO for image identification and classification for underwater scenes. The proposed methodology, developed for such an application, modifies the cell size in EY5, which prepares the model for applying BDO for higher speed and accuracy. The BDO algorithm in the EY5 comprises initialization, moving simulation, foraging, communication and cooperation, pursuit of prey, adaptation, learning procedures, and the termination procedure. A number of experiments were performed to evaluate the potentials of the developed optimized EY5 against competitors such as DeepSeaNet and MCANet. A true positive rate of 61.11% with a true negative rate of 58.82%, precision at 70% and has high overall classification accuracy. The improvement of the EY5 model is called as BeardYOLO Here, it shows more efficiency in terms of clustering and classifying the underwater images and becomes one of the best solutions to work with underwater images. The study thereby raises the possibility of achieving superior performance in a specific field by integrating both the theories of advanced optimization with modern artificial neural networks.