YOLO11-YX: An efficient algorithm for marine debris target detection.

The escalating issue of marine debris pollution has positioned automated detection as a pivotal technology in combating this environmental challenge. Traditional object detection methodologies, however, often grapple with limitations in accuracy and robustness, particularly in complex backgrounds and with small objects. Addressing these limitations, this paper introduces YOLO11-YX, an enhanced algorithm derived from YOLO11s, integrating three novel modules: the SDown downsampling module, the C3SE feature extraction module, and the FAN feature fusion module. The SDown module amalgamates feature information from various image regions during downsampling, effectively reducing data dimensionality and computational complexity while preserving intricate details. The C3SE module refines the conventional YOLO architecture by streamlining the convolutional structure and embedding SENet, thereby ameliorating the redundancy in multi-layer bottlenecks and bolstering detection efficacy in intricate environments. The FAN module augments the network's capacity to discern image details and contextual information at its terminus, accentuating the features of diminutive targets and elevating detection precision. Empirical results demonstrate that YOLO11-YX surpasses YOLO11s by achieving a 2.44% enhancement in detection accuracy for marine debris, culminating in a 62.32% accuracy rate. This advancement furnishes a potent and dependable solution for the automated detection of marine debris, heralding extensive applicability.