Hi , I want to know how to build a condition network ?
My idea is :
Condition is a label ,so I want to use one-hot input and fc_layer in condition network.
Main network is full convolution network with 4 HaarDownsampling.
This is failed ,but condtion and featuredon't agree : [(512,) ] vs [(1, 64, 64)]
I think the output of condition network must be same shape as main network feature?
I can understand colorization_minial_example experiments code it is the same as paper said, but colorization_cINN, have many new functions ,so Can you write a blog about it?
Hi , I want to know how to build a condition network ?
My idea is :
This is failed ,but condtion and feature
don't agree : [(512,) ] vs [(1, 64, 64)]I think the output of condition network must be same shape as main network feature?
I can understand colorization_minial_example experiments code it is the same as paper said, but colorization_cINN, have many new functions ,so Can you write a blog about it?