AdderNet Problem

[MillionLee]

Hello，I want to realize the Transposed Adder layer according to the relationship between convolution and transposed convolution, but from the result I found that the image it generates has very obvious checkerboard artifacts. I think that it is because the variance of the output of AdderNet is much larger than CNN. So I want to know if you could give me some suggestions to solve it? Thanks very much!!!

[ranery]

Interesting. I haven't tried transposed adder layer.

But that may be related to super-resolution tasks. You can check the AdderSR paper below to see whether they have encountered similar issues.

https://openaccess.thecvf.com/content/CVPR2021/papers/Song_AdderSR_Towards_Energy_Efficient_Image_Super-Resolution_CVPR_2021_paper.pdf