Conversion of trained Neural Network into C++ data/code using MATLAB Coder.

I have trained a neural network for my specific purpose to detect different types of arrow using a Webcam using Transfer Learning Method. I have used AlexNet for transfer Learning. The trained network is stored as an array of Convolutinal Neural Network [25*1]. The trained network is saved as a variable in MATLAB workspace and loaded in the corresponding MATLAB code for classification of arrows.

Now, i am trying to integerate this feature with a Robotino mobile Roboter and to do that i have to create a new function block in Robotinoview 2 software.

The new function block in RobotinoView 2 can only be created or defined using C++ program. So I want to convert the MATLAB Code into C++ code. I understood how to convert a MATLAB function into a C++ code.

In order to classify the types of arrows, the code has to be loaded with the trained Network array. But I dont understand how can i convert the trained convolutional neural network int to a C++ code or to a C++object/class.

Answer 1

[Mukund Sankaran]

You may find these resources helpful:

https://www.mathworks.com/help/deeplearning/deep-learning-code-generation.html

https://www.mathworks.com/help/deeplearning/examples/code-generation-for-deep-learning-networks.html (contains an example for generating C++ code for image classification using Alexnet)

[Philip Kanjirathingal Kurian]

Thank you for the help. It was helpful to understand more about code conversion using deep learning.