About the expansion of the pre-training dataset

[dasimaduoshuaia]

Dear Author： Thank you so much for this exciting work!

We note that you have collected a large amount of optical remote sensing images to pre-train the diffusion model. In order to explore the migration capability of your proposed method on multi-source remote sensing data, we would like to further expand this dataset with SAR images and retrain the diffusion model, so we take the liberty to ask you if you can provide us with your pre-training dataset.

Gratefully thanks for your consideration, and look forward your reply.

Dr.Wu

[wgcban]

Hi @dasimaduoshuaia Thank you! We collected remote sensing images from publicly available RGB remote sensing datasets: LEVIR-CD (train-set), WHU-CD (train-set), DSIFN-CD (train-set), SeCo-100k images, and BigEarthNet images. Please download all the images from those dataset and dump inside a one folder prior to starting the diffusion training. Download links to some datasets can be found in my other repos like ChangeFormer, DDPM-CD, etc.

[dasimaduoshuaia]

Dear Author：

It's great honor to receive your reply informing us of the composition of the pre-trained dataset！If there is any latest development, we will share with you in time.

Thank you again for your information.