Regarding "WSIPenMarkingRemoval"

[shubhangi-9595]

Hi Mr. Jiang,

Your method looks very interesting to me. I have a few Whole Slide Images that I stitched together using their FOVs, so I already have access to the individual FOVs as well.

From what I understand, your model requires both marked and clean image pairs for training. Unfortunately, I don't have clean versions of my test data. However, I was thinking of generating synthetic training data by adding marker annotations on top of my clean FOVs and training the model that way.

Before I proceed, I wanted to ask if you already have a pretrained model available that I could use directly on my data. If not, I’ll go ahead with creating the training dataset myself as described above.

Also, I noticed that staining intensity varies across my WSIs, and I’m concerned this might affect performance, especially if the model is trained on lighter-stained images and tested on darker ones. Have you encountered this issue, and do you have any recommendations for handling it?

Thanks!