search

swz30 / MPRNet

[CVPR 2021] Multi-Stage Progressive Image Restoration. SOTA results for Image deblurring, deraining, and denoising.
Other
1.18k stars 192 forks source link

Train on dual_pixels_dataset_canon? #34

Closed pingjun18-li closed 3 years ago

pingjun18-li commented 3 years ago

Hello, thank you for your contribution. I want to know have you tried to use dual pixel images to train your network? for example dual_pixels_dataset_canon, which only contain 500 pairs dual pixel labels. if yes, can you share the codes or ideas about how use it on that projects?

swz30 commented 3 years ago

Hi @pingjun18-li

We did try it by participating in NTIRE 2021 Defocus Deblurring using Dual-pixel Images Challenge. We concatenate the left and right images and provide them as input to the network. Just change in_c=6

https://github.com/swz30/MPRNet/blob/435f483b6433a6cf8d43f4df9c20eebba47587bb/Deblurring/MPRNet.py#L239

Our method ranked second. In fact, the winning solution also uses our supervised attention module (SAM). The challenge report will appear as a workshop paper in CVPR2021. We will notify you once it becomes publicly available.

pingjun18-li commented 3 years ago

Congratulations on your good grades again. And I will try to train on that data , looking forward to your project publicly available also.