How to input image features to the featureinputLayer of a deep neural network in MATLAB.

5 次查看(过去 30 天)
I have built a CNN model using transfer learning with DeepNetworkDesigner app. I have used Vgg-16 and replaced its layers such that the 'fc6' output is fed to a concatenation layer that concatenates the 'fc6' output and the output of 'featureInputLayer'. When I analyze the network, I am getting multiple errors. Some of them were resolved but some still persist and make the network invalid.
I wanted to implement this network to train on images as well as on image features. I have 2000 feature vectors each of size 1x20160 which I want to feed at featureinputLayer. What data transformations are needed to get this... Please guide!
Photos displaying data, network and errors are attached here:

回答(1 个)

ziyang huang
ziyang huang 2022-8-9
openExample('nnet/CreateDlnetworkWithTwoInputLayersExample')
Try to copy this command to command window to see the built-in example, I've seen this example by chance which is almost the same as your situation and I think might help you. Seems it needs to create a dlnetwork object or whatever.

类别

Help CenterFile Exchange 中查找有关 Image Data Workflows 的更多信息

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by