Support for n-dimensional Deep Neural Network

[John Puskin]

Hi Davis,
As far as I see, dlib supports only 2-d input images in DNN architectures. What about 1-d or 3-d, 4-d signals/images?

[Davis]

The tensors are nominally 4D. But you can add whatever additional annotations you want to them in the annotation() field. The tensor is just a block of memory and each layer interprets them however that layer does. So there is no limitation on what you do with it.

[John Puskin]

We are very glad to see that if you can add a new example about this issue to dlib/example in the upcoming versions.