Add BRIGHT dataset

[nilsleh]

This PR adds the BRIGHT dataset.

Dataset Features:

* Pre-disaster optical images from MAXAR, NAIP, NOAA Digital Coast Raster Datasets, and the National Plan for Aerial Orthophotography Spain
* Post-disaster SAR images from Capella Space and Umbra
* high image resolution of 0.3-1m

Dataset Format:

* Images are in GeoTIFF format with pixel dimensions of 1024x1024
* Pre-disaster are three channel images
* Post-disaster SAR images are single channel but repeated to have 3 channels

@ChenHongruixuan @olidietrich Thank you for the nice work and making the dataset public. I was wondering whether you could include the split .txt files found here inside the zip file on Huggingface such that everything is in one place? And of course if you have any other comments about the PR, feel free to let us know below.

[adamjstewart]

Hmm, does this make sense? Is this only because they want to use RGB models pre-trained on ImageNet?

[nilsleh]

I think now that we are implementing this as the "Challenge" version of the dataset it's okay to do?

[adamjstewart]

It depends. If we want to use one of our S1 pretrained models, not duplicating may make more sense. I would personally leave it up to the user to add a transform if they want this.

[adamjstewart]

I think I'm still against this. It would prevent us from using our S1 pretrained models on S1 imagery.

[ChenHongruixuan]

This PR adds the BRIGHT dataset.

Dataset Features:

* Pre-disaster optical images from MAXAR, NAIP, NOAA Digital Coast Raster Datasets, and the National Plan for Aerial Orthophotography Spain
* Post-disaster SAR images from Capella Space and Umbra
* high image resolution of 0.3-1m

Dataset Format:

* Images are in GeoTIFF format with pixel dimensions of 1024x1024
* Pre-disaster are three channel images
* Post-disaster SAR images are single channel but repeated to have 3 channels

@ChenHongruixuan @olidietrich Thank you for the nice work and making the dataset public. I was wondering whether you could include the split .txt files found here inside the zip file on Huggingface such that everything is in one place? And of course if you have any other comments about the PR, feel free to let us know below.

Hi @nilsleh ,

Thank you very much for your contribution and efforts in integrating BRIGHT into torchgeo! We are a little hesitant to include the txt file you mentioned into this for torchgeo, as that is only for the DFC25 division. That's why we provide these txt files in github instead of original dataset zip file, especially considering that the current BRIGHT is not final version.

Best,

[nilsleh]

Thank you very much for your contribution and efforts in integrating BRIGHT into torchgeo! We are a little hesitant to include the txt file you mentioned into this for torchgeo, as that is only for the DFC25 division. That's why we provide these txt files in github instead of original dataset zip file, especially considering that the current BRIGHT is not final version.

Thanks for getting back @ChenHongruixuan , no problem, we can also host them on our torchgeo Hugginface, just to make downloading more easier. Also very much looking forward to the final version of the dataset. Feel free to let me know when that is openly available, such that we can integrate support for that as well here.

[ChenHongruixuan]

Thank you very much for your contribution and efforts in integrating BRIGHT into torchgeo! We are a little hesitant to include the txt file you mentioned into this for torchgeo, as that is only for the DFC25 division. That's why we provide these txt files in github instead of original dataset zip file, especially considering that the current BRIGHT is not final version.

Thanks for getting back @ChenHongruixuan , no problem, we can also host them on our torchgeo Hugginface, just to make downloading more easier. Also very much looking forward to the final version of the dataset. Feel free to let me know when that is openly available, such that we can integrate support for that as well here.

@nilsleh , thank you so much for your support!!