Initial figure feedback 3-Figures.ipynb

[dshean]

https://github.com/uw-cryo/DeepDEM/blob/main/notebooks/3-Figures.ipynb

The hillshade should use cmap='gray' not default viridis.

We need a figure that shows all input layers for examples, including triangulation error and NDVI grids.

Consider using matplotlib-scalebar(https://github.com/ppinard/matplotlib-scalebar) - simpler than current implementation, but whatever works is fine.

Remove masking of vegetation in ASP DSM - LIDAR DSM, masking out vegetated regions figure after fixing DSM idw vs. max bug (which will improve agreement over dense vegetation). Create additional panel for "DeepDEM DSM - LiDAR DSM" running inference on all tiles, which should hopefully show improvement.

All figures using RdBu color ramp should use black background, as white is used to represent valid values near 0.

Final Plots showing initial ASP DEM should not use filled version. There is no need for a separate panel showing nodata mask if unfilled version is used. Can potentially include a supplementary figure showing unfilled vs filled ASP DEM, or arrange subplots to show in these multi-panel figures, with clear explanation of what we start with vs. what goes into the model. I think a separate figures showing pre-processing steps, maybe flow chart is going to be best way to discuss, as we do lots of other preprocessing (including the co-registration and alignment).

[dshean]

compressed() vs flatten() for histograms