u-net question

[ILoveSorasakiHina]

Dear Thomas Capelle,

First of all, thank you for your experiment on MoveMNIST. It has been very helpful to me.

However, I have a small question.

As I understand it, MoveMNIST is a spatiotemporal sequence prediction problem. UNet is typically used for semantic segmentation, where multiple channels represent different semantic categories.

In your implementation, you use multiple channels to represent multiple time steps, which I find a bit difficult to understand.

Once again, I truly appreciate your work—it has been very beneficial to me.

Best regards,

ILoveTakanashiHoshino