Automatic Ship Object Detection Model Based on YOLOv4 with Transformer Mechanism in Remote Sensing Images

Despite significant advancements in object detection technology, most existing networks fail to investigate global aspects while extracting features from the inputs and cannot automatically adjust based on characteristics of inputs. The present study addresses this problem by proposing a network consisting three stages: preattention, attention, prediction. In preattention stage, framework is selected images’ objects. attention transformer structure introduced. Taking into account target, combines self-attention module model convolution operation integrate image local for detection, thus improving ship target accuracy. This uses mathematical methods obtain results predictive testing prediction stage. above improvements are You Only Look Once version 4 (YOLOv4) framework, named “Auto-T-YOLO”. achieves highest accuracy 96.3% SAR Ship Detection dataset (SSDD) compared other state-of-the-art (SOTA) model. It 98.33% 91.78% offshore inshore scenes, respectively. experimental verify practicality, validity, robustness proposed