Questions about your released model/checkpoint

[Vandermode]

Hi, I find the checkpoints/models released in this repo is a bit tricky for me since it is not mentioned in the paper.

In your paper, you apply the trained model to HSI denoising under different noise setting. By default, I would assume you use a single trained model to tackle all different noise setting.

But in your repo, you release two models, i.e. HSID-CNN_Realdata_iter_1000000 and HSID-CNN_noiselevel100_iter_600000.

So I was wondering, is that means you train multiple models to deal with each case respectively?
What's the model "HSID-CNN_Realdata_iter_1000000" resort to? How is that trained?

Thank you for your attention and looking forward to your reply.

[qzhang95]

1. In the simulated experiments, HSID-CNN trains multiple models to deal with each case, respectively.
2. In the real-data experiments, HSID-CNN trains single models with random noise for different bands. Current models employs [0-75] strength noise.

[qzhang95]

Actually, you can also train the single model for different data or noise strengh through adding random noise like DnCNN, SRMD, etc.