Image upsampling with fractional stride

[jumerckx]

I'm trying to implement a image detection network wich needs upsampling layers. I've read that an efficient way of doing upsampling is by using a convolutional layer with a fractional stride. Is there a way to implement this in Flux?
Any chance someone could implement an upsample layer like in other frameworks such as pytorch https://pytorch.org/docs/stable/_modules/torch/nn/modules/upsampling.html ?

Thanks in advance,
Jules

[avik-pal]

The linked layer in Pytorch is not the same as what you are looking for. As per your description, you are talking about ConvTranspose which is performed by Fractionally Strided Convolutions.
The ConvTranspose is currently in a WIP PR #311.
For simple NearestUpsampling you can just use the Base.repeat function.

[jumerckx]

Oh, thanks!