how is the in-line googlenet or inception v3 created using codes?

Hi, I need to create an NIN structure just like googlenet.

so how to fulfill it in deep leaning toolbox? ( not with Deep Network Designer but with codes)

Answer 1

[Vishal Bhutani]

By my understanding it seems that you want to create a similar architecture to googlenet.

You can refer to the link attached below for refering the googlenet architecture and you can use MATLAB function for creating the architecture:

Googlenet architecture: https://www.analyticsvidhya.com/blog/2018/10/understanding-inception-network-from-scratch/

MATLAB functions which will be helpful to create the architecture:

• https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layer.imageinputlayer.html

• https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layer.convolution2dlayer.html

• https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layer.batchnormalizationlayer.html

• https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layer.relulayer.html

• https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layer.fullyconnectedlayer.html

• https://www.mathworks.com/help/deeplearning/ref/nnet.cnn.layer.softmaxlayer.html

• https://www.mathworks.com/help/deeplearning/ref/classificationlayer.html

Also you can try using googlenet inbuilt architecture in MATLAB as a pre-trained network and train it for your dataset.

Hope it helps.

[Jack Xiao]

thank you, maybe imitating the keras sripts is a good choice. but I really hope there can be scripts of constructing these typical models as every deep learning platform has that scripts.