FoveaBox: Beyond Anchor-based Object Detector论文笔记

FoveaBox: Beyond Anchor-based Object Detector
arXiv:1904.03797v1 [cs.CV] 8 Apr 2019
论文地址：
(https://www.researchgate.net/publication/332300345_FoveaBox_Beyond_Anchor-based_Object_Detector)
提出了一种新的不需要anchor(锚框)的目标检测方法，采用resnet-101的foceaBox的AP达到了42.1
1.Introduction

3.FoveaBox
FoveaBox的对象检测

3.1.FeaturePyramidNetworkBackbone
FoveaBox is a single, unified network composed of a backbone network and two task-specific subnetworks. The backbone is responsible for computing a convolutional feature map over an entire input image and is an off-the-shelf convolutional network. The first subnet performs per pixel classification on the backbone’s output; the second subnet performs bounding box prediction for the corresponding position.
FoveaBox是一个单一的、统一的网络，由一个主干网络和两个特定于任务的子网络组成。主干负责计算整个输入图像上的卷积特征图，是一个现成的卷积网络。第一子网对主干的输出按像素进行分类;第二个子网络对相应的位置执行包围框预测。

We adopt t