PatrickTUM/SEN12MS-CR-TS ? reverse-engineered prompt

Build me a Python project for cloud removal on satellite imagery using the SEN12MS CR TS dataset. I want to be able to load the data, train a model, and run tests from the command line, with support for multi temporal optical and radar inputs, cloud covered scenes, and cloud free targets.

Please make it easy to point at a local dataset folder, choose train or test split, and run a basic baseline training and evaluation flow. It should include a working data loader I can use on its own, plus simple scripts for training and testing a model. If there are useful defaults for Sentinel 1 and Sentinel 2 inputs, use them, and keep the setup friendly for someone who just wants to get results quickly. Use sensible paths, save checkpoints and outputs, and include clear examples of how to launch everything. If anything about the dataset format matters, please wire that in and make it robust. Look up current docs online if you need to.