Improving Dataset Handling for Sentinel-1 and Sentinel-2 Images

#1
by rwilliams - opened

This pull request addresses challenges in handling Sentinel-1 and Sentinel-2 image data, specifically focusing on images that span a 5090 x 5090 meter area at a resolution of 10 meters per pixel. The default 512 x 512 pixel size causes issues when fed into a model because of the blank pixels. Additionally, strategies for mitigating the effects of padding, as well as improvements in data split and re-composition, are proposed.

thank you!

csaybar changed pull request status to merged

Sign up or log in to comment