Improve SD image quality and reduce artefacts without any additional training, simply by reweighting skip connections in the decoder stage of a diffusion Unet decoder
i know i should read the paper but... what does reweighting the skip connections mean in this case? like... how?
In the decoder, the features from the unet blocks get concatenated with features from the encoder layer through 'skip connections'. The paper discusses how rescaling the backbone features (element-wise multiplication by some scalar) before concatenation improves image quality.