Underwater object detection using global flower pollination-based sunflower optimization (GFP-SFO) on YOLOv6

Concealing the mystery of the oceans is a tedious process,ineluctable for many applications, and a hot research topic in the world. With advanced technology, enormous data has been gathered using underwater sensors, posing challenges in investigating data with blurred, low-resolution images of inadequate quality. This might result in arduous training and difficulty in exploring uncontrolled objects. Processing these images manually and with machine learning algorithms might incur a substantialcost and time and be prone to errors. To overcome these issues, we propose a novel YOLOv6-based global flower pollination-based sunflower optimization (GFP-SFO) algorithm for underwater object detection. The existing YOLOv6 algorithm detects underwater objects such as corals, urchins, rocks, and fish from the images, whose detection accuracy can be improved with the proposed GFP-SFO algorithm. The images are taken from the EUVP dataset and are pre-processed by the Volterra filter. The pre-processed images are fed into the YOLOv6 for object detection. The simulations are carried out in Python software. The performance of the work is compared with state-of-art research in terms are accuracy, sensitivity, specificity, recall, and precision. The comprehensive study shows that the proposed work outperforms the existing literature.