yangxy/PASD

Missing Comparison with SUPIR [CVPR 24]

Closed this issue · 2 comments

lyf1212 commented

Thank you for your wonderful work! I notice that you don't provide a comparison with SUPIR [CVPR24]1, but your work utilize the base backbone of ControlNet and SUPIR also, and you two works both modify the mid-connection of ControlNet, in your work is "PACD" and in SUPIR is "ZeroSFT block", so I am really wonder more quantitative comparison between your method and SUPIR if it convenient for your teams.

Footnotes

  1. https://github.com/Fanghua-Yu/SUPIR ↩

yangxy commented

PASD predates SUPIR. Please refer to the SUPIR's paper for the comparison results.

lyf1212 commented

Sorry for my careless ignorance. Thank you for your reply and this work!