When I want to upsample features in the middle of a model, usually by applying deconvolution. However, does it make sense that adopting upsample in network?
When I want to upsample features in the middle of a model, usually by applying deconvolution. However, does it make sense that adopting upsample in network?