backropagation for categorical cross entropy loss with sigmoid output layer

Question

I am trying to implement CNN in C++. I have a sigmoid output layer.To implement categorical cross entropy loss, I am following this. Am I correct in assuming that the only diffe