Litcius/Paper detail

YOLO-Ship: A Visible Light Ship Detection Method

S. R. Zhou, Jun Yin

20222022 2nd International Conference on Consumer Electronics and Computer Engineering (ICCECE)28 citationsDOI

Abstract

YOLO detection model is a lightweight and efficient target detection algorithm. For realizing real-time accurate recognition of visible ship targets, an improved ship target detection method YOLO-Ship derived from YOLOv5 is advanced. This model uses MixConv to improve the traditional convolution operation and coordinated attention mechanism, and uses Focal Loss and CIoU Loss to optimize loss functions of the original method. On Ship7000 dataset, YOLO-Ship detection method has a more outstanding accuracy than YOLOv5 and YOLOv3 algorithms, and its average detection accuracy under different IoU thresholds is improved from 0.514 to 0.728, which can effectively meet the practical application requirements.

Topics & Concepts

Computer scienceConvolution (computer science)Artificial intelligenceObject detectionComputer visionPattern recognition (psychology)Artificial neural networkAdvanced Neural Network ApplicationsInfrared Target Detection MethodologiesVisual Attention and Saliency Detection
YOLO-Ship: A Visible Light Ship Detection Method | Litcius