I have a 2D point cloud data time series. The shape of the data is [N, T]. N is a big number（millions) and T is the number of 2D point cloud images and generally less than 300. I also have the coordinates of these N points with shape of [N, 2] (the longitudes and the latitudes). I want to interactively visualize these data in a map.
Specifically, I want: 1.) interactively zoom in or out to show both the specific region and the big picture; 2.) I can click one point and plot the time series of this point; 3.) I can switch which image to be shown among the T images.
I know it is impossible and unnecessary to plot that much of points. A dynamic aggregation and rasterization would be a better choice. https://egms.land.copernicus.eu/ is a very good example. It meets all of my requirement. It only shows all points when zoom into a small region. Otherwise, only a small number of points are shown. It could be the reason of its efficiency.
The only similar technique I found in Python ecosystem is Holoviews+Datashader. But I find it is still not very efficient. The reason could be that datashader rasterizes the point cloud on the fly, but that website rasterizes the point cloud previously and only on several specific zoom level.
I want to know if there are any python packages that can visualize such large dataset efficiently, or if there are any better ways to utilize holoviews and datashader. Thanks!