I have a following neural network architecture:
A(1x16x5x5)---\\ -> concat(A,B) (1x32x5x5) --> Conv2D -> classification B(1x16x5x5)