作者:Xue YangJirui YangJunchi YanYue ZhangTengfei ZhangZhi GuoSun XianKun fu

机构:

论文全称:Dynamic Anchor Learning for Arbitrary-Oriented Object Detection

**发布时间:**2020

发布期刊:arxiv

论文地址:https://arxiv.org/pdf/2012.04150v2.pdf

代码:

https://github.com/ming71/DAL

https://github.com/ming71/UCAS-AOD-benchmark

创新:

一、摘要

Arbitrary-oriented objects widely appear in natural scenes, aerial photographs, remote sensing images, etc., and thus arbitrary-oriented object detection has received considerable attention. Many current rotation detectors use plenty of anchors with different orientations to achieve spatial alignment with ground truth boxes. Intersection-over-Union (IoU) is then applied to sample the positive and negative candidates for training. However, we observe that the selected positive anchors cannot always ensure accurate detections after regression, while some negative samples can achieve accurate localization. It indicates that the quality assessment of anchors through IoU is not appropriate, and this further leads to inconsistency between classification confidence and localization accuracy. In this paper, we propose a dynamic anchor learning (DAL) method, which utilizes the newly defined matching degree to comprehensively evaluate the localization potential of the anchors and carries out a more efficient label assignment process. In this way, the detector can dynamically select high-quality anchors to achieve accurate object detection, and the divergence between classification and regression will be alleviated. With the newly introduced DAL, we can achieve superior detection performance for arbitrary-oriented objects with only a few horizontal preset anchors. Experimental results on three remote sensing datasets HRSC2016, DOTA, UCAS-AOD as well as a scene text dataset ICDAR 2015 show that our method achieves substantial improvement compared with the baseline model. Besides, our approach is also universal for object detection using horizontal bound box. The code and models are available at https://github.com/ming71/DAL.

Untitled

二、研究背景

目前,基于anchor的算法在训练时首先根据将预设的anchor和目标根据IoU大小进行空间匹配,以一定的阈值(如0.5)选出合适数目的anchor作为正样本用于回归分配的物体。但是这会导致两个问题:

  1. **进一步加剧的正负样本不平衡:**对于旋转目标检测而言,预设旋转anchor要额外引入角度先验,使得预设的anchor数目成倍增加。此外,旋转anchor角度稍微偏离gt会导致IoU急剧下降,所以预设的角度参数很多。(例如旋转文本检测RRD设置13个角度,RRPN每个位置54个anchor)。
  2. **分类回归的不一致:**即预测结果的分类得分和定位精度不一致,导致通过NMS阶段以及根据分类conf选检测结果的时候有可能选出定位不准的,而抑制了定位好的anchor得到的预测框

为了进一步验证上述第二点现象是否具有普遍性,统计了训练过程的所有样本IoU分布图,以及分类回归分数散点图

  1. 由下图(a),(b)可知,74%左右的正样本anchor回归的pred box后依然是高质量样本(IoU>0.5);58%的高质量样本回归自负样本,即负样本还有很大的利用空间,也就是说目前所使用的正负样本分配方法(基于IoU的分配方法)并不合理
  2. 而图(c)说明的**是当前的基于输入IoU的标签分配会诱导分类分数和anchor初始定位能力成正相关,而我们期望的结果是pred box的分类回归能力成正相关,图(d)说明的是使用的分配策略和我们的期望结果并不符合,并且可以认为基于输入IoU的标签分配是导致分类回归不一致的根源之一,**简单的理解就是:划分样本的时候指定的初始对齐很好的为正样本,其回归后就算产生了不好的预测结果,分类置信还是很高,因为分类回归任务是解耦的;反之很多初始对齐不好的anchor被分成负样本,即使能预测好,由于分类置信度很低,无法在inference时被输出。

综上所述,定位性能并不完全依赖于anchor和GT之间的空间对齐(一般用两者之间的IoU来衡量)