How to develop a CNN classification with meta, train and test data

I am currenlty working with the GTSRB dataset, I need to build an algorithm for pattern recognition to classify the different traffic signs using the DL-Toolobox, I was planing to train a CNN. I need to develop this algorithm in a script, but also use the Simulink block as well. This includes having a a phase of training, validation and testing. I do not know how to begin, can anyone guide me with a basic example or recommend me some specific documentation that could be helpful?

Answer 1

[Mahesh Taparia]

Hi

You can use pretrained network already exist in DL toolbox and finetune the model on your data. You can refer this documentation on the list of pretrained network. For custom network creation and training, you can refer this documentation and trained the network on your data. After performing the training, you can test it in MATLAB as well as in simulink. To do this in simulink, you can use predict block which takes the pretrained network as input from the workspace/ specified folder. For more information, you can refer this documentation.

Hope it will help!