mwieland / s1s2_water Goto Github PK

Hello,

thank you very much for providing this very nice dataset ! I am sure there has been a tremendous amount of work behind this.

I just have one question : I would like to use s1s2_water as a training dataset and apply the model to other images (like Copernicus freely avalable data), so I would like to know what pre-processing steps you performed on your images. Especially for the s1 images.

In your paper you write "The Sentinel-1 images are geometrically corrected and radiometrically calibrated according to the method described by Twele et al. [33]." And [33] states « After data pre-processing using the S1TBX, the calibrated data are filtered to reduce SAR inherent speckle noise. » Since there are many de-speckling filters implemented in the S1TBX (ESA Sentinel-1 toolbox), can I ask wich one you chose and with which parameters ?

Thanks again for releasing this dataset which will allow great progress in water bodies detection !

All the best

Hello again !

I downloaded your dataset and, unless i'm mistaken, it seems that scenes 31 and 59 do not have the corresponding s1 images, s1 masks and s1 valid masks.
Maybe you removed these 2 for a reason ?
(but scene 31 was in your test set for sentinel1, so I guess at least this one has a matching s1 image :) )

Thanks a lot for your help !

Best regards,
Laetitia

Thank you for this excellent work! Will the train and test code, along with the dataloader, be released? It would be very helpful for us to utilize the dataset!