1.May I ask is it only necessary to provide the restored image during the testing phase, as in the verification phase, or should we also provide the running test code?
2.Where are the weight coefficients of the evaluation indicators？
Same question. How about large files like pretrained Stable Diffusion model weights. Shall we provide a download URL or directly call it in our code (assuming that the pretrained model weights will be automatically downloaded from huggingface)
Posted by: duanxiongwa @ March 14, 2025, 4:04 p.m.Q: May I ask is it only necessary to provide the restored image during the testing phase, as in the verification phase, or should we also provide the running test code?
A; In CodaLab, the participants should submit their restored images to update the TEST leaderboard. Note that the results in CodaLab is just for reference , and it will NOT be seen as official submission.
In Final Submission (i.e. Fact sheets and code/executable submission period), the participants should submit their codebase, with the template following https://github.com/zhengchen1999/NTIRE2025_RealWorld_Face_Restoration.
Q: How about large files like pretrained Stable Diffusion model weights. Shall we provide a download URL or directly call it in our code (assuming that the pretrained model weights will be automatically downloaded from huggingface)?
A: You can assume that the pretrained model weights will be automatically downloaded from huggingface. As for other links, the participants can use some cloud drive and share the links.
Q: Where are the weight coefficients of the evaluation indicators？
A: We have already updated the coefficients in the GitHub repo (https://github.com/zhengchen1999/NTIRE2025_RealWorld_Face_Restoration). Thanks for your patience.