Run on custom data

[aping-2193]

I tried to train your method on my own data, and I have got some meaningful results. But unfortunately, the default parameters are not the best choice for every scenario, so I'd like to ask what hyperparameters I can tune to get the best results for different data? Thanks for sharing your experience. @akshatdave @yyiz

[akshatdave]

Hi, Thank you for your interest.
You can refer to Appendix B on Implementation Details in our arxiv version ( https://arxiv.org/pdf/2203.13458.pdf ) for important hyperparameters required for training.

[junxiaofanchen]

Hello, I would like to ask how to use polarized RAW images, get specular and diffuse reflection and normal vector separation of EXR files. Is there any part of your code that is written? I want to be able to collect my own datasets to test and work with your excellent algorithms

[junxiaofanchen]

I tried to train your method on my own data, and I have got some meaningful results. But unfortunately, the default parameters are not the best choice for every scenario, so I'd like to ask what hyperparameters I can tune to get the best results for different data? Thanks for sharing your experience. @akshatdave @yyiz

Have you implemented your own dataset operation? How do you separate specular and diffuse reflection from surface normals in your collected data set? Is it possible to leave a contact information for communication?