面向小型 Logo 检测的跨方向任务解耦网络（英文版）

英文标题：A Cross-direction Task Decoupling Network for Small Logo Detection中文摘要：本文提出了一种名为 Cross-direction Task Decoupling Network (CTDNet) 的小型 LOGO 检测算法，使用 Cross-direction Feature Pyramid (CFP) 实现横向和纵向特征融合，并提出了 Multi-frequency Task Decoupling Head (MTDH) 来解决分类和定位任务，使用多频注意力卷积实现更准确的回归。通过实验在四个 LOGO 数据集上验证了该方法的有效性和效率。英文摘要：Logo detection plays an integral role in many applications. However, handlingsmall logos is still difficult since they occupy too few pixels in the image,which burdens the extraction o