20180711_Deep Neural Networks for Object Detection_基于DNN的对象检测

Deep Neural Networks for Object Detection用于物体检测的深度深度神经网络

（点击标题链接原文http://papers.nips.cc/paper/5207-deep-neural-networks-for-object-detection.pdf）

---

1、Introduction介绍

DNNs与传统分类方法区别

1. First, they are deep
   architectures which have the capacity to learn more complex models than shallow ones [2].深度模型架构。
2. This
   expressivity and robust training algorithms allow for learning powerful object representations without the need to hand design features.不需要手工设计特征。

In this paper, we exploit the power of DNNs for the problem of object detection, where we not only
classify but also try to precisely localize objects.这篇论文解决了物体检测的分类问题和精确定位对象。
We present a formulation which is capable of predicting the bounding boxes of multiple objects in
a given image. 能够预测多个物体的边界框。

Figure 1: A schematic view of object detection as DNN-based regression.图1:作为基于DNN的回归的对象检测的原理图。

Figure 2: After regressing to object masks across several scales and large image boxes, we perform
object box extraction. The obtained boxes are refined by repeating the same procedure on the sub
images, cropped via the current object boxes. For brevity, we display only the full object mask,
however, we use all five object masks.图2：在几个刻度和大图像框上回归到对象蒙版后，我们执行
对象框提取。 通过在子上重复相同的过程来细化所获得的框
图像，通过当前对象框裁剪。 为简洁起见，我们只显示完整的对象蒙版，
但是，我们使用所有五个对象蒙版。

In this paper, we demonstrate that DNN-based regression is capable of learning features which
are not only good for classification, but also capture strong geometric information. 证明：分类，捕获几何信息。

2、Related Work相关工作

物体检测研究最广泛的范例之一the deformable part-based model可变性部件的模型，pictorial structure.图案结构It can be considered as a 2-layer model – parts being
the first layer and the star model being the second layer.2层模型，部分是第一层，星形是第二层。DNNs图层是通用的。

Deep architectures for object detection and parsing have been motivated by part-based models and
traditionally are called compositional models，用于对象检测和解析的深层体系结构受到基于部件的模型的推动传统上称为组合模型，where the object is expressed as layered composition
of image primitives.其中对象表示为分层组合
图像基元。

3、DNN-based Detection基于DNN的探测

The core of our approach is a DNN-based regression towards an object mask核心是基于DNN的对象掩码回归。（图1）

4、Detection as DNN Regression DNN回归检测

Our network is based on the convolutional DNN defined by [14]. It consists of total 7 layers, the
first 5 of which being convolutional and the last 2 fully connected. Each layer uses a rectified linear
unit as a non-linear transformation. Three of the convolutional layers have in addition max pooling.
For further details, we refer the reader to [14]我们的网络基于[14]定义的卷积DNN。 它由总共7层组成其中前5个是卷积的，后2个是完全连接的。 每层使用整流线性单位作为非线性变换。 其中三个卷积层还有最大池。有关详细信息，请参阅[14]

Θ are the parameters of the network and N is the total number of pixels. Θ是网络的参数，N是像素的总数。

The network is trained by minimizing the L2 error for predicting a ground truth mask m 2 [0; 1]N
for an image x:通过最小化L2误差来训练网络以预测地面实况掩模m 2 [0;1] n的对于图像x：

where the sum ranges over a training set D of images containing bounding boxed objects which are
represented as binary masks其中总和范围超过包含边界框对象的图像的训练集D.表示为二进制掩码

5、Precise Object Localization via DNN-generated Masks基于DNN生成掩模的精确目标定位

论文这部分只要对三个具有挑战性的问题进行分析和解决：

First, a single object mask might not be sufficient to disambiguate objects
which are placed next to each other.第一，单个对象掩码可能不足以消除彼此相邻的对象的歧义。

Second, due to the limits in the output size, we generate masks
that are much smaller than the size of the original image.第二，由于输出大小的限制，生成掩码比原始图像尺寸小得多。（would be insufficient to
precisely localize an object不足以精确定位一个对象）。

Finally, since we use as an input the full
image, small objects will affect very few input neurons and thus will be hard to recognize.第三，输入整张图片，小物体对输入神经元的影响很少，导致很难识别。

In the
following, we explain how we address these issues.下面将解释如何解决这些问题。

5.1、Multiple Masks for Robust Localization 用于稳健本地化的多个掩码

mh,h∈{ full,bottom,top,left,right} m h