gy65896 / onerestore Goto Github PK

Dear @gy65896

It's really nice work and thanks for making the dataset and code public!

I am curious if you have tested your method on another all-in-one IR setting, i.e., the Prompt/AirNet setting.

If you have tried the experiments, may I know the performance?

Best regards and many thanks,

Hi @gy65896

Thanks for your nice work.

I am trying to reproduce your method from scratch.

I would like to know what GPU the text/visual embedder is trained on.

Best regards and many thanks,

Hi @gy65896

Thanks for your nice work.

I am trying to reproduce your method from scratch.

However, when I train the proposed method (i.e., embedded) with 4 GPUs, there is always only one GPU is used though I already used 'CUAD_VISIBLE_DEVICES=0,1,2,3'

It would be super helpful to hear some suggestions from you.

Best regards and many thanks,

嗨，首先祝贺你们做出如此优秀的工作！也很感激你们的开源！在阅读完你们精彩的论文之后，我心中产生了一个疑问渴望得到你们的解答。
我注意到，BID（ECCV 2022）在Task II.A 数据集上的实验与你们的论文方向十分匹配，请问一下你们有在 Task II.A 数据集上做过相关的实验吗？

Hello, thanks for your work.

Could you please provide code for how to synthesize low-quality images?