Improve 3D volume renderer

[forman]

The current, experimental 3D volume renderer is already useful, but has several limitations and problems. Here is a list of things that would improve user experience and performance..

• Exploit existing dataset zoom levels, currently
  • users can only view small amounts of data, users forced to create subsets of data by AOI
  • users cannot know in advance if selected data subset is small enough
  • users cannot visualize global coverage data for higher resolution data (Gunnar’s idea: use suitable lower pyramid level)
• Fix bug in computation of normals where data intersects the volume boundaries.
• Fix issue where 3D canvas does not adjust size on resize of right panel.
• Show labels and directions for x, y, z axes.
• Show currently selected time as highlighted x/y-plane along time dimension.
• Support perspective projection. We currently use orthographic due to a problem in ray casting.
• Recognize NaNs in value. Currently NaN is set to zero, because WebGL textures cannot have NaNs.
• Mask volume by selected place’s shape i.e. data subsets should be created according polygon boundaries. We currently use a polygon bounding box.
• Improve rendering quality of ISO surface: voxel interpolation, material, lightning.
• Reduce amount of transferred data if dtype is 8 or 16 bit integer. Currently we always use dtype float32.
• Add i18, add translations for DE and SE, currently we only have EN only

[gunbra32]

some even more sophisticated features:

• move time slider through 3D cube and visualise corresponding 2D image in the other window
• show selected features in Cube, e.g. reference points or shape files delineating AOIs, ...

[forman]

I just the wonderful package react-three-fiber. Should definitely use it when addressing this issue:

• Docs: https://docs.pmnd.rs/react-three-fiber/getting-started/introduction
• GH: https://github.com/pmndrs/react-three-fiber