Use a Quad Tree approach to create adaptive bin sizes and compute a histogram (counts or density) in 2 dimensions. Can specify limit on number of points per bin, and how many levels in the tree to go down. Uses np.histogram2d to lazily build each level of the tree, may be slow for very large number of points, or N levels.
Some simple experiments with using a triangular mesh included as well, but abandoned for now since KDTree & Ball Tree ideas already exist that are probably better suited for many other cases.
This project was developed in part at the online.tess.science meeting, which took place globally in 2020 September.
Code originally developed for the EBHRD project