I make UNet model and try to solve this competition like a one-shot segmentation problem, but the score is too low (~0.11 public lb). This is unstable model with insufficient amount of data. But it may be interesting and useful for someone else. There are also some 3D (x, y, elevation) interactive plots.
https://www.kaggle.com/simakov/zindi-flood-in-malawi-unet-baseline
Thanks a lot skittles, for this amazing kernel. Even though its not useful for this competition at this stage, there's still a lot to learn using this.
Thanks a lot for this great