How to decide the number of input nodes, hidden layers and its nodes, output nodes?

for my understanding, input layers equal to number of features (without target feature, and in my project will be 4 features)

and the number of output nodes based on the number of values of target feature (3 values in my target feature)

with 10,000 instances ( 80% in Training - 10% in Evaluation - 10% in Testing)

and I have no Idea about how to decide the number of hidden layers and nodes

and I'm not sure if I had a right understand or not