Convolution Neural Network configuration -
how can choose cnn architecture? example, have n classes , images of size m*m, how can determine layer sizes , count should use? there universal formula? in case have 16k classes , images of size 64*64. can me it?
there no techniques determine layer sizes , count should use. it's more of trial , error. there no universal formula. choose layer size , train network. change , train again, till satisfied accuracy.
Comments
Post a Comment