一种高效的多尺度目标无锚框检测方法
An efficient multi-scale anchor-free object detection method
随着深度学习在目标检测领域的发展，有越来越多类型的高精度检测算法涌现。然而，现有方法对动态目标的检测准确率较低。针对上述问题，本文提出一种高效的多尺度目标无锚框检测方法，该方法是一种无锚检测器，不需定义显式的锚框。本文方法将不同尺度的特征进行融合，并在骨干网络之后添加了特征金字塔和可变形卷积，以准确检测不同尺寸的物体。仿真分析与实验结果表明，本文算法可以极大地提高动态目标检测的精度。
计算技术、计算机技术
全卷积关键点检测动态目标检测准确率
杨松祥,沈全成,贾爽,张玉波.一种高效的多尺度目标无锚框检测方法[EB/OL].(2025-10-31)[2025-11-02].http://www.paper.edu.cn/releasepaper/content/202510-39.点此复制
With the development of deep learning in the field of object detection, an increasing number of high-precision detection algorithms have emerged. However, existing methods exhibit lower detection accuracy for dynamic objects. To address this issue, this paper proposes an efficient multi-scale anchor-free object detection method. This approach does not require the explicit definition of anchor boxes. The proposed method integrates features at different scales and incorporates a feature pyramid and deformable convolution layers after the backbone network to accurately detect objects of various sizes. Simulation analysis and experimental results demonstrate that the proposed algorithm significantly improves the accuracy of dynamic object detection.
fully convolutionalkeypoint detectiondynamic objectsdetection accuracy
展开英文信息
评论