learning
/
DeepLearning-500-questions
mirror of https://github.com/scutan90/DeepLearning-500-questions.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Update 第四章_经典网络.md 

inception images and referance
This commit is contained in:
CoderOverflow 2019-04-03 11:28:20 +08:00 committed by GitHub
parent abf7bdb8ca
commit 52c6f3df88
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 6 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
ch04_经典网络/第四章_经典网络.md
View File

@ -236,7 +236,7 @@ GoogLeNet作为2014年ILSVRC在分类任务上的冠军，以6.65%的错误率
​ 如图4.9中所示，GoogLeNet相比于以前的卷积神经网络结构，除了在深度上进行了延伸，还对网络的宽度进行了扩展，整个网络由许多块状子网络的堆叠而成，这个子网络构成了Inception结构。图4.9为Inception的四个版本：$Inception_{v1}$在同一层中采用不同的卷积核，并对卷积结果进行合并;$Inception_{v2}$组合不同卷积核的堆叠形式，并对卷积结果进行合并;$Inception_{v3}$则在$v_2$基础上进行深度组合的尝试;$Inception_{v4}​$结构相比于前面的版本更加复杂，子网络中嵌套着子网络。 ​ 如图4.9中所示，GoogLeNet相比于以前的卷积神经网络结构，除了在深度上进行了延伸，还对网络的宽度进行了扩展，整个网络由许多块状子网络的堆叠而成，这个子网络构成了Inception结构。图4.9为Inception的四个版本：$Inception_{v1}$在同一层中采用不同的卷积核，并对卷积结果进行合并;$Inception_{v2}$组合不同卷积核的堆叠形式，并对卷积结果进行合并;$Inception_{v3}$则在$v_2$基础上进行深度组合的尝试;$Inception_{v4}​$结构相比于前面的版本更加复杂，子网络中嵌套着子网络。
![](./img/ch4/img_inception_02.png) ![](./img/ch4/img_inception_02.png)
![](./img/ch4/img_inception_05.png)![](./img/ch4/img_inception_06.png)![](./img/ch4/img_inception_07.png) <img src="./img/ch4/img_inception_05.png" width="250" height="250" /><img src="./img/ch4/img_inception_06.png" width="240" height="250" /><img src="./img/ch4/img_inception_07.png" width="270" height="250" />
![](./img/ch4/img_inception_03.png)![](./img/ch4/img_inception_04.png) ![](./img/ch4/img_inception_03.png)![](./img/ch4/img_inception_04.png)
​ 图 4.10 Inception$_{v1-4}$结构图 ​ 图 4.10 Inception$_{v1-4}$结构图
@ -286,5 +286,9 @@ GoogLeNet作为2014年ILSVRC在分类任务上的冠军，以6.65%的错误率
[8] Christian Szegedy, Sergey Ioffe, Vincent Vanhoucke, Alex Alemi. [Inception-v4, Inception-ResNet and [8] Christian Szegedy, Sergey Ioffe, Vincent Vanhoucke, Alex Alemi. [Inception-v4, Inception-ResNet and
the Impact of Residual Connections on Learning](https://arxiv.org/pdf/1602.07261.pdf), 2016. the Impact of Residual Connections on Learning](https://arxiv.org/pdf/1602.07261.pdf), 2016.
[9] Sik-Ho Tsang. [review-inception-v4-evolved-from-googlenet-merged-with-resnet-idea-image-classification](https://towardsdatascience.com/review-inception-v4-evolved-from-googlenet-merged-with-resnet-idea-image-classification-5e8c339d18bc), 2018 [9] Sik-Ho Tsang. [review-inception-v4-evolved-from-googlenet-merged-with-resnet-idea-image-classification](https://towardsdatascience.com/review-inception-v4-evolved-from-googlenet-merged-with-resnet-idea-image-classification-5e8c339d18bc), 2018.
[10] Zbigniew Wojna, Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jonathon Shlens. [Rethinking the Inception Architecture for Computer Vision](https://arxiv.org/pdf/1512.00567v3.pdf), 2015.
[11] Christian Szegedy, Wei Liu, Yangqing Jia, Pierre Sermanet, Scott Reed, Dragomir Anguelov, Dumitru Erhan, Vincent Vanhoucke, Andrew Rabinovich. [Going deeper with convolutions](https://arxiv.org/pdf/1409.4842v1.pdf), 2014.