how can i input two inputs to the deep network, one from the imageinputlayer and the other from featureinputlayer

Question

Nagwa megahed 2022-3-4

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1663829-how-can-i-input-two-inputs-to-the-deep-network-one-from-the-imageinputlayer-and-the-other-from-feat

评论： Mahesh Taparia 2022-3-10

how can i input the images and the extracted features by one of the algorithms to a deep network

this example shows how can i build the structure of the network with an input layer and feature input layer

but how i can train the network with the two inputs???

2 个评论
显示无隐藏无

yanqi liu 2022-3-7

what is two inputs mean? two dimension or two variable？

Nagwa megahed 2022-3-8

two variables ( one is an image datastore and the other is an a table (or array ) with set of features that represent each image)

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Mahesh Taparia 2022-3-8

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1663829-how-can-i-input-two-inputs-to-the-deep-network-one-from-the-imageinputlayer-and-the-other-from-feat#answer_913114

Hi

You can refer to this example which shows how to combine image and features and then pass it to a CNN. It create a dlnet object which you can train it using custom training loop. For more information, you can refer to this documentation. Hope it will help!

2 个评论
显示无隐藏无

Nagwa megahed 2022-3-8

Dear sir, thanx for your answer

but i already build the network and combined the input layer with a feature input layer but i ask how can i train the network with two inputs , to train a network you should determine a single input , this input may be an image datastore or image features

my question is how to combine two inputs(image+features) in a single input or if there is another way to choose two inputs for the network

Mahesh Taparia 2022-3-10

Hi

You can concatenate the image and features in the channels. For example if you have images of size MxNxZ and features of size MxNxK, then concatenate into mxnx(Z+K) size and pass in to network.

请先登录，再进行评论。