How do I Verify Robustness of Deep Learning Neural Network with custom data?

I want to repurpose method given here for the custom dataset. I have a folder full of greyscale images with foldernames as theire respective labels. And I have a dlnetwork pretrained and ready to be used.

From what I understand I will have to convert my dataset to usable form. The exampe given uses 'digitTest4DArrayData'. My question is how do I convert my data in the said format? Or is there any way I can use what I have and still make it work?

Answer 1

[Ranjeet]

verifyNetworkRobustness uses dlarray as inputs for lower and upper limits. As shown in the example, the input images have been converted to dlarray before feeding into verifyNetworkRobustness function.

Based on the type of data, different dlarray data formats are possible as mentioned in the following documentation –

https://www.mathworks.com/help/deeplearning/ref/dlarray.html#mw_e439e5de-77e9-4399-a2e8-2cc62d2ab575

It is advised to first decide on what data format is suitable, and then convert the input to dlarray.