I wanted to implement a CNN but explore what happens when my first layer is Fully Connected. I still want to use convolutions of course but I wanted to apply them layer. I noticed that the input then losses its 3D structure. Does that mean I can only apply 1d convolutions after that? Is there a non-trivial way to recover the 3d structure so that 2d convolutions may be applied?

Hopefully when I reconstruct it to have 3d structure the 3d structure is somehow meaningful...

Cross posted:

• https://forums.fast.ai/t/how-do-i-recover-the-3d-structure-of-a-layer-after-a-fully-connected-layer-or-a-flatten-layer/52489
• https://discuss.pytorch.org/t/how-do-i-recover-the-3d-structure-of-a-layer-after-a-fully-connected-layer-or-a-flatten-layer/53313
• How do I recover the 3D structure of a layer after a fully-connected layer?

I want to implement a CNN, but I want to explore what happens when my first layer is a fully-connected one. I still want to use convolutions, of course, but I want to apply them after the first layer. I noticed that the input then loses its 3D structure. Does that mean I can only apply 1d convolutions after that? Is there a non-trivial way to recover the 3d structure, so that 2d convolutions may be applied?

Hopefully, when I reconstruct it to have 3d structure the 3d structure is somehow meaningful.

I also posted this question at https://forums.fast.ai/t/how-do-i-recover-the-3d-structure-of-a-layer-after-a-fully-connected-layer-or-a-flatten-layer/52489 and https://discuss.pytorch.org/t/how-do-i-recover-the-3d-structure-of-a-layer-after-a-fully-connected-layer-or-a-flatten-layer/53313.