Currently the package is made with the assumption that it will be used on cells that have already been identified as grid cells. A natural next step would be to implement a grid score method in the Cell class, and implement a test identifying grid cells based on some grid score criteria, shuffling etc.

Another thing to consider implementing is kernel smoothed density estimation in the CellCollection class, to inform the clustering of cells into modules (e.g. to determine the n_clusters parameter for K-means clustering).

The package has grown to a considerable size, and should get its own test suite. The ad-hoc method of writing notebooks and looking at the output is no longer sufficient.

Having a bunch of plotting methods on the objects is making them heavy and bloated. All plotting functionality should be factored out into a separate module with plotting functions taking as arguments instances of classes for which plotting is relevant. This should be a flexible and general-purpose module, not an opinionated, seaborn-dependent and option-deprived module like 'niceplots'.

The docstrings are poorly formatted and should be rewritten according to the numpy docstring standard. But... ain't nobody got time for that!