Is it possible to use the convolutional functions of TensorFlow with my own neural network?

Question

That is, a convolutional network consists of its layers of convolution, pooling, etc. And in the end a fully connected network. Well, TF provides all the tools, I want it to be that instead of using the TF network only to use what goes before and the result of that use it in a totally own NETWORK. Is it possible? Could you help me?

Answer 1

等待大神解答