Extracting (ranked) softmax values for each validation image

Hi everyone,

I trained a model (fine tuning) to classify 10 types of images. I was just wondering if there was a simple way to return, say, a matrix containing all validation images (with their respective names/labels) and their predictive scores (classification confidence) ?

Thank you !

Best regards.

Answer 1

[Srivardhan Gadila]

Use the activations function to get the output of softmaxLayer & use the max function to get the maximum of all scores i.e., score of the predicted class. Also I think you can use the same Name-Value Pair Arguments & Syntax used for predict function. You can refer to Visualize Activations of a Convolutional Neural Network for more examples on the usage of activations function.