search

tyshiwo / FSRNet

Demo code for our CVPR'18 paper "FSRNet: End-to-End Learning Face Super-Resolution with Facial Priors" (SPOTLIGHT Presentation)
255 stars 53 forks source link

Is a universal model suitable for all LR faces possible? #1

Closed XiangyuWu closed 6 years ago

XiangyuWu commented 6 years ago

Hi, tyshiwo!

Thanks for your great work. When implementing your project, it confused me that why different models are needed for different dataset? As shown in your paper, the only input requested for this network in the test stage is just a LR image. Are there great differences between Helen dataset and celeba dataset, for example, the crop strategy, which lead to different models needed? And I test helen dataset with celeba model and the contray, it turned out that different models for different datasets are indeed needed.

tyshiwo commented 6 years ago

Hi, thanks for your interests on our work.

Different datasets may have different distributions, which requires different models for the best performance. However, if different datasets share similar distribution, one model may be enough.

In our case, the celeba model may be better, since it is trained by a larger dataset.

XiangyuWu commented 6 years ago

You are quite right, and sorry that I made a mistake just now...