Improving Neural Network classification accuracy

I'm trying to classify images using NN classifier available in Matlab R2019a.

I have 1800 images into 3 classes (Hard, Semi, Soft).

I have 68 features (descriptors).

When I classify into 2 classes (hard or soft): I get 93% accuracy

When I classify into 3 classes (hard -- semi -- soft): I get 75% accuracy

Question is: how to improve the accuracy for the 3 classes model?

Any ideas?

P.S. I can always send a sample data file if needed.

[Mohammad Sami]

You will need to try various NN models, to try and improve your accuracy.

Also try all the models in the classification learner app.