深度卷积网络中的空间金字塔池化用于视觉识别
需积分: 9 121 浏览量
更新于2024-09-09
收藏 3.97MB PDF 举报
"这篇论文提出了一种在深度卷积神经网络（CNN）中用于视觉识别的空间金字塔池化方法，称为SPP-Net（Spatial Pyramid Pooling Network）。它解决了传统CNN需要固定尺寸输入图像的问题，提高了对任意大小或尺度图像的识别准确性，并且对物体变形具有鲁棒性。"
在深度学习领域，卷积神经网络（CNN）已经成为图像识别和计算机视觉任务的核心技术。然而，传统的CNN架构通常需要固定尺寸的输入图像，这在实际应用中可能会限制模型的泛化能力，特别是对于不同尺寸或比例的对象。论文"Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition"由Kaiming He、Xiangyu Zhang、Shaoqing Ren和Jian Sun提出，旨在解决这一问题。
SPP-Net的主要创新是引入了空间金字塔池化层。这一层能够对不同大小的特征图进行池化操作，生成固定长度的表示，从而摆脱了对输入图像尺寸的依赖。空间金字塔池化通过在多个层次上进行池化，模仿了空间金字塔模型，这样可以捕获不同尺度的信息，对尺度变化有较好的适应性。
具体来说，空间金字塔池化将特征图分割成多个子区域，每个子区域执行最大池化操作，然后将所有子区域的池化结果拼接成一个固定长度的向量。这种方法允许网络接收不同大小的输入，并且在训练过程中保持网络参数不变，这对于图像分类和其他视觉任务尤其有用。
在ImageNet 2012数据集上的实验显示，SPP-Net显著提高了多种已发表的CNN架构的识别精度，无论这些架构的设计如何。同时，在Pascal VOC 2007和Caltech 101数据集上，SPP-Net仅使用单个全图像表示并在没有微调的情况下达到了最先进的分类结果。这表明SPP-Net的强大性能，尤其是在处理具有复杂尺度变化和物体变形的场景时。
SPP-Net的另一个优点是它可以轻松地插入到现有的CNN架构中，无需对整个网络进行重新设计。这使得它成为现有CNN模型的一种有效增强手段，对于提升模型的泛化能力和鲁棒性具有重要意义。此外，SPP-Net对于实时应用，如目标检测和图像分割，也有着显著的影响，因为它可以减少对输入图像预处理的需求，加快计算速度。
"Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition"是深度学习领域的一个重要里程碑，它推动了CNN在处理变尺寸输入和处理复杂场景方面的进步，为后续的研究和应用奠定了坚实的基础。
3
ImageNet datasets. But the feature computation in R-
CNN is time-consuming, because it repeatedly applies
the deep convolutional networks to the raw pixels
of thousands of warped regions per image. In this
paper, we show that we can run the convolutional
layers only once on the entire image (regardless of
the number of windows), and then extract features
by SPP-net on the feature maps. This method yields
a speedup of over one hundred times over R-CNN.
Note that training/running a detector on the feature
maps (rather than image regions) is actually a more
popular idea [23], [24], [20], [5]. But SPP-net inherits
the power of the deep CNN feature maps and also the
flexibility of SPP on arbitrary window sizes, which
leads to outstanding accuracy and efficiency. In our
experiment, the SPP-net-based system (built upon the
R-CNN pipeline) computes convolutional features 30-
170× faster than R-CNN, and is overall 24-64× faster,
while has better or comparable accuracy. We further
propose a simple model combination method to boost
the result on the Pascal VOC 2007 detection task.
A preliminary version of this manuscript has been
published in ECCV 2014 [25]. Based on [25], we
attended the competition of ILSVRC 2014 [26], and
ranked #2 in object detection and #3 in image clas-
sification (both are provided-data-only tracks) among
all 38 teams. There are a few modifications made
over [25] for ILSVRC 2014. We show that the SPP-
nets can boost various networks that are deeper and
larger (Sec. 3.1.2-3.1.4) over the no-SPP counterparts.
Further, driven by our detection framework, we find
that multi-view testing on feature maps with flexibly
located/sized windows (Sec. 3.1.5) can increase the
classification accuracy. This manuscript also provides
the details of these modifications.
2 DEEP NETWORKS WITH SPATIAL PYRA-
MID POOLING
2.1 Convolutional Layers and Feature Maps
Consider the popular seven-layer architectures [3], [4].
The first five layers are convolutional, some of which
are followed by pooling layers. These pooling layers
can also be considered as “convolutional”, in the sense
that they are using sliding windows. The last two
layers are fully connected, with an N-way softmax as
the output, where N is the number of categories.
The deep network described above needs a fixed
image size. However, we notice that the requirement
of fixed sizes is only due to the fully-connected layers
that demand fixed-length vectors as inputs. On the
other hand, the convolutional layers accept inputs of
arbitrary sizes. The convolutional layers use sliding
filters, and their outputs have roughly the same aspect
ratio as the inputs. These outputs are known as feature
maps [1] - they involve not only the strength of the
responses, but also their spatial positions.
In Figure 2, we visualize some feature maps. They
are generated by some filters of the conv
5
layer
(the fifth convolutional layer). Figure 2(c) shows the
strongest activated images of these filters in the Ima-
geNet dataset. We see a filter can be activated by some
semantic content. For example, the 55-th filter (Fig-
ure 2, bottom left) is most activated by a circle shape;
the 66-th filter (Figure 2, top right) is most activated
by a ∧-shape; and the 118-th filter (Figure 2, bottom
right) is most activated by a ∨-shape. These shapes
in the input images (Figure 2(a)) activate the feature
maps at the corresponding positions (the arrows in
Figure 2).
It is worth noticing that we generate the feature
maps in Figure 2 without fixing the input size. These
feature maps generated by deep convolutional lay-
ers are analogous to the feature maps in traditional
methods [27], [28]. In those methods, SIFT vectors
[29] or image patches [28] are densely extracted and
then encoded, e.g., by vector quantization [16], [15],
[30], sparse coding [17], [18], or Fisher kernels [19].
These encoded features consist of the feature maps,
and are then pooled by Bag-of-Words (BoW) [16] or
spatial pyramids [14], [15]. Analogously, the deep
convolutional features can be pooled in a similar way.
2.2 The Spatial Pyramid Pooling Layer
The convolutional layers accept arbitrary input sizes,
but they produce outputs of variable sizes. The classi-
fiers (SVM/softmax) or fully-connected layers require
fixed-length vectors. Such vectors can be generated
by the Bag-of-Words (BoW) approach [16] that pools
the features together. Spatial pyramid pooling [14],
[15] improves BoW in that it can maintain spatial
information by pooling in local spatial bins. These
spatial bins have sizes proportional to the image size,
so the number of bins is fixed regardless of the image
size. This is in contrast to the sliding window pooling
of the previous deep networks [3], where the number
of sliding windows depends on the input size.
To adopt the deep network for images of arbi-
trary sizes, we replace the last pooling layer (e.g.,
pool
5
, after the last convolutional layer) with a spatial
pyramid pooling layer. Figure 3 illustrates our method.
In each spatial bin, we pool the responses of each
filter (throughout this paper we use max pooling).
The outputs of the spatial pyramid pooling are kM-
dimensional vectors with the number of bins denoted
as M (k is the number of filters in the last convo-
lutional layer). The fixed-dimensional vectors are the
input to the fully-connected layer.
With spatial pyramid pooling, the input image can
be of any sizes. This not only allows arbitrary aspect
ratios, but also allows arbitrary scales. We can resize
the input image to any scale (e.g., min(w, h)=180, 224,
...) and apply the same deep network. When the
input image is at different scales, the network (with
剩余13页未读，继续阅读
2024-07-20 上传
2024-07-24 上传
2024-07-23 上传
2023-03-16 上传
2023-01-25 上传
点击了解资源详情
点击了解资源详情
点击了解资源详情
2024-07-22 上传
YELLOWPIPLE
- 粉丝: 0
- 资源: 3
我的内容管理
展开
最新资源
- Angular程序高效加载与展示海量Excel数据技巧
- Argos客户端开发流程及Vue配置指南
- 基于源码的PHP Webshell审查工具介绍
- Mina任务部署Rpush教程与实践指南
- 密歇根大学主题新标签页壁纸与多功能扩展
- Golang编程入门：基础代码学习教程
- Aplysia吸引子分析MATLAB代码套件解读
- 程序性竞争问题解决实践指南
- lyra: Rust语言实现的特征提取POC功能
- Chrome扩展：NBA全明星新标签壁纸
- 探索通用Lisp用户空间文件系统clufs_0.7
- dheap: Haxe实现的高效D-ary堆算法
- 利用BladeRF实现简易VNA频率响应分析工具
- 深度解析Amazon SQS在C#中的应用实践
- 正义联盟计划管理系统：udemy-heroes-demo-09
- JavaScript语法jsonpointer替代实现介绍
