采用注意力机制与改进YOLOv5的光伏用地检测

陈笛, 彭秋志, 黄培依, 刘雅璇

Detecting land for photovoltaic development based on the attention mechanism and improved YOLOv5

CHEN Di, PENG Qiuzhi, HUANG Peiyi, LIU Yaxuan

表1 改进算法平均精度区别

Tab.1 Mean average precision of improved algorithm

算法 加权双向 特征金字塔 Ghost 卷积 协同注意 力机制 mAP/% YOLOv5 × × × 80.25 算法1 √ × × 81.73 算法2 √ √ × 82.62 YOLOv5-pv √ √ √ 84.68