VERYDEEPCONVOLUTIONALNETWORKSFORLARGE-SCALEIMAGERECOGNITION

VGG-NET

需积分: 47 177 浏览量更新于2023-03-16 评论 1 收藏 195KB PDF 举报

身份认证购VIP最低享 7 折!

领优惠券(最高得80元）

资源详情

资源评论

资源推荐

arXiv:1409.1556v6 [cs.CV] 10 Apr 2015

Published as a conference paper at ICLR 2015

VERY DEEP CONVOLUTIONAL NETWORKS

FOR LARGE-SCALE IMAGE RECOGNITION

Karen Simonyan

∗

& Andrew Zisserman

Visual Geometry Group, Department of Engineering Science, University of Oxford

{karen,az}@robots.ox.ac.uk

ABSTRACT

In this work we investigate the effect of the convolutional network depth on its

accuracy in the large-scale image recognition setting. Our main contribution is

a thorough evaluation of networks of increasing depth using an architecture with

very small (3 × 3) convolution ﬁlters, which shows that a signiﬁcant improvement

on the prior-art conﬁgurations can be achieved by pushing the depth to 16–19

weight layers. These ﬁndings were the basis of our ImageNet Challenge 2014

submission, where our team secured the ﬁrst and the second places in the localisa-

tion and classiﬁcation tracks respectively. We also show that our representations

generalise well to other datasets, where they achieve state-of-the-art results. We

have made our two best-performing ConvNet models publicly available to facili-

tate further research on the use of deep visual representations in computer vision.

1 INTRODUCTION

Convolutional networks (ConvNets) have recently enjoyed a great success in large-scale im-

age and video recognition (Krizhevsky et al., 2012; Zeiler & Fergus, 2013; Sermanet et al., 2014;

Simonyan & Zisserman, 2014) which has become possible due to the large public image reposito-

ries, such as ImageNet (Deng et al., 2009), and high-performancecomputing systems, such as GPUs

or large-scale distributed clusters (Dean et al., 2012). In particular, an important role in the advance

of deep visual recognition architectures has been played by the ImageNet Large-Scale Visual Recog-

nition Challenge (ILSVRC) (Russakovsky et al., 2014), which has served as a testbed for a few

generations of large-scale image classiﬁcation systems, from high-dimensional shallow feature en-

codings (Perronnin et al., 2010) (the winner of ILSVRC-2011) to deep ConvNets (Krizhevsky et al.,

2012) (the winner of ILSVRC-2012).

With ConvNets becoming more of a commodity in the computer vision ﬁeld, a number of at-

tempts have been made to improve the original architecture of Krizhevsky et al. (2012) in a

bid to achieve better accuracy. For instance, the best-performing submissions to the ILSVRC-

2013 (Zeiler & Fergus, 2013; Sermanet et al., 2014) utilised smaller receptive window size and

smaller stride of the ﬁrst convolutional layer. Another line of improvements dealt with training

and testing the networks densely over the whole image and over multiple scales (Sermanet et al.,

2014; Howard, 2014). In this paper, we address another important aspect of ConvNet architecture

design – its depth. To this end, we ﬁx other parameters of the architecture, and steadily increase the

depth of the network by adding more convolutional layers, which is feasible due to the use of very

small (3 × 3) convolution ﬁlters in all layers.

As a result, we come up with signiﬁcantly more accurate ConvNet architectures, which not only

achieve the state-of-the-art accuracy on ILSVRC classiﬁcation and localisation tasks, but are also

applicable to other image recognition datasets, where they achieve excellent performance even when

used as a part of a relatively simple pipelines (e.g. deep features classiﬁed by a linear SVM without

ﬁne-tuning). We have released our two best-performing models

to facilitate further research.

The rest of the paper is organised as follows. In Sect. 2, we describe our ConvNet conﬁgurations.

The details of the image classiﬁcation training and evaluation are then presented in Sect. 3, and the

∗

current afﬁliation: Google DeepMind

current afﬁliation: University of Oxford and Google DeepMind

http://www.robots.ox.ac.uk/

vgg/research/very_deep/

Published as a conference paper at ICLR 2015

conﬁgurations are compared on the ILSVRC classiﬁcation task in Sect. 4. Sect. 5 concludes the

paper. For completeness, we also describe and assess our ILSVRC-2014 object localisation system

in Appendix A, and discuss the generalisation of very deep features to other datasets in Appendix B.

Finally, Appendix C contains the list of major paper revisions.

2 CONVNET CONFIGURATIONS

To measure the improvement brought by the increased ConvNet depth in a fair setting, all our

ConvNet layer conﬁgurations are designed using the same principles, inspired by Ciresan et al.

(2011); Krizhevsky et al. (2012). In this section, we ﬁrst describe a generic layout of our ConvNet

conﬁgurations (Sect. 2.1) and then detail the speciﬁc conﬁgurations used in the evaluation (Sect. 2.2).

Our design choices are then discussed and compared to the prior art in Sect. 2.3.

2.1 ARCHITECTURE

During training, the input to our ConvNets is a ﬁxed-size 224 × 224 RGB image. The only pre-

processing we do is subtracting the mean RGB value, computed on the training set, from each pixel.

The image is passed through a stack of convolutional (conv.) layers, where we use ﬁlters with a very

small receptive ﬁeld: 3 × 3 (which is the smallest size to capture the notion of left/right, up/down,

center). In one of the conﬁgurations we also utilise 1 × 1 convolution ﬁlters, which can be seen as

a linear transformation of the input channels (followed by non-linearity). The convolution stride is

ﬁxed to 1 pixel; the spatial padding of conv. layer input is such that the spatial resolution is preserved

after convolution, i.e. the padding is 1 pixel for 3 × 3 conv. layers. Spatial pooling is carried out by

ﬁve max-pooling layers, which follow some of the conv. layers (not all the conv. layers are followed

by max-pooling). Max-pooling is performed over a 2 × 2 pixel window, with stride 2.

A stack of convolutional layers (which has a different depth in different architectures) is followed by

three Fully-Connected (FC) layers: the ﬁrst two have 4096 channels each, the third performs 1000-

way ILSVRC classiﬁcation and thus contains 1000 channels (one for each class). The ﬁnal layer is

the soft-max layer. The conﬁguration of the fully connected layers is the same in all networks.

All hidden layers are equipped with the rectiﬁcation (ReLU (Krizhevsky et al., 2012)) non-linearity.

We note that none of our networks (except for one) contain Local Response Normalisation

(LRN) normalisation (Krizhevsky et al., 2012): as will be shown in Sect. 4, such normalisation

does not improve the performance on the ILSVRC dataset, but leads to increased memory con-

sumption and computation time. Where applicable, the parameters for the LRN layer are those

of (Krizhevsky et al., 2012).

2.2 CONFIGURATIONS

The ConvNet conﬁgurations, evaluated in this paper, are outlined in Table 1, one per column. In

the following we will refer to the nets by their names (A–E). All conﬁgurations follow the generic

design presented in Sect. 2.1, and differ only in the depth: from 11 weight layers in the network A

(8 conv. and 3 FC layers) to 19 weight layers in the network E (16 conv. and 3 FC layers). The width

of conv. layers (the number of channels) is rather small, starting from 64 in the ﬁrst layer and then

increasing by a factor of 2 after each max-pooling layer, until it reaches 512.

In Table 2 we report the number of parameters for each conﬁguration. In spite of a large depth, the

number of weights in our nets is not greater than the number of weights in a more shallow net with

larger conv. layer widths and receptive ﬁelds (144M weights in (Sermanet et al., 2014)).

2.3 DISCUSSION

Our ConvNet conﬁgurations are quite different from the ones used in the top-performing entries

of the ILSVRC-2012 (Krizhevsky et al., 2012) and ILSVRC-2013 competitions (Zeiler & Fergus,

2013; Sermanet et al., 2014). Rather than using relatively large receptive ﬁelds in the ﬁrst conv. lay-

ers (e.g. 11 × 11 with stride 4 in (Krizhevsky et al., 2012), or 7 × 7 with stride 2 in (Zeiler & Fergus,

2013; Sermanet et al., 2014)), we use very small 3 × 3 receptive ﬁelds throughout the whole net,

which are convolved with the input at every pixel (with stride 1). It is easy to see that a stack of two

3 × 3 conv. layers (without spatial pooling in between) has an effective receptive ﬁeld of 5 × 5; three

Published as a conference paper at ICLR 2015

Table 1: ConvNet conﬁgurations (shown in columns). The depth of the conﬁgurations increases

from the left (A) to the right (E), as more layers are added (the added layers are shown in bold). The

convolutional layer parameters are denoted as “convhreceptive ﬁeld sizei-hnumber of channelsi”.

The ReLU activation function is not shown for brevity.

ConvNet Conﬁguration

A A-LRN B C D E

11 weight 11 weight 13 weight 16 weight 16 weight 19 weight

layers layers layers layers layers layers

input (224 × 224 RGB image)

conv3-64 conv3-64 conv3-64 conv3-64 conv3-64 conv3-64

LRN conv3-64 conv3-64 conv3-64 conv3-64

maxpool

conv3-128 conv3-128 conv3-128 conv3-128 conv3-128 conv3-128

conv3-128 conv3-128 conv3-128 conv3-128

maxpool

conv3-256 conv3-256 conv3-256 conv3-256 conv3-256 conv3-256

conv1-256 conv3-256 conv3-256

conv3-256

maxpool

conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512

conv1-512 conv3-512 conv3-512

conv3-512

maxpool

conv3-512 conv3-512 conv3-512 conv3-512 conv3-512 conv3-512

conv1-512 conv3-512 conv3-512

conv3-512

maxpool

FC-4096

FC-1000

soft-max

Table 2: Number of parameters (in millions).

Network A,A-LRN B C D E

Number of parameters 133 133 134 138 144

such layers have a 7 × 7 effective receptive ﬁeld. So what have we gained by using, for instance, a

stack of three 3 × 3 conv. layers instead of a single 7 × 7 layer? First, we incorporate three non-linear

rectiﬁcation layers instead of a single one, which makes the decision function more discriminative.

Second, we decrease the number of parameters: assuming that both the input and the output of a

three-layer 3 × 3 convolution stack has C channels, the stack is parametrised by 3





= 27C

weights; at the same time, a single 7 × 7 conv. layer would require 7

= 49C

parameters, i.e.

81% more. This can be seen as imposing a regularisation on the 7 × 7 conv. ﬁlters, forcing them to

have a decomposition through the 3 × 3 ﬁlters (with non-linearity injected in between).

The incorporation of 1 × 1 conv. layers (conﬁguration C, Table 1) is a way to increase the non-

linearity of the decision function without affecting the receptive ﬁelds of the conv. layers. Even

though in our case the 1 × 1 convolution is essentially a linear projection onto the space of the same

dimensionality (the number of input and output channels is the same), an additional non-linearity is

introduced by the rectiﬁcation function. It should be noted that 1 × 1 conv. layers have recently been

utilised in the “Network in Network” architecture of Lin et al. (2014).

Small-size convolution ﬁlters have been previously used by Ciresan et al. (2011), but their nets

are signiﬁcantly less deep than ours, and they did not evaluate on the large-scale ILSVRC

dataset. Goodfellow et al. (2014) applied deep ConvNets (11 weight layers) to the task of

street number recognition, and showed that the increased depth led to better performance.

GoogLeNet (Szegedy et al., 2014), a top-performing entry of the ILSVRC-2014 classiﬁcation task,

was developed independently of our work, but is similar in that it is based on very deep ConvNets

剩余13页未读，继续阅读

autocyz

粉丝: 533
资源: 7

会员权益专享

VERY DEEP CONVOLUTIONAL NETWORKS FOR LARGE-SCALE IMAGE RECOGNITI...

评论0

会员权益专享

最新资源

VERY DEEP CONVOLUTIONAL NETWORKS FOR LARGE-SCALE IMAGE RECOGNITI...

评论0

【5】Very deep convolutional networks for large-scale image recognition.pdf

VERY DEEP CONVOLUTIONAL NETWORKS.pdf

十种流行网络在cifar-10数据集上的应用

基于 tensorflow+VGG-16 的图像分类识别算法

very deep convolutional networks for large-scale image recognition

VGG16网络英文全程

请给我找几篇有关神经网络的文献

卷积神经网络反向传播有没有推荐的论文

推荐通过CNN提取图像语义特征的论文

卷积神经网络经典论文

卷积神经网络图像分类的参考文献

图像分类的特征提取可以参考哪篇文献？

计算机毕设英文参考文献

基于神经网络的花卉识别系统设计与实现任务书

为卷积神经网络的研究现状写一些参考文献

国内图像识别分类研究现状和成果

有哪些关于卷积神经网络研究现状的参考文献

卷积神经网络参考文献

定义VGGNet网络结构

卷积神经网络分类器应用

会员权益专享

最新资源