Qu- want to downsample image_url plots

Hi, I am planning on plotting a lot of image_urls which, if I plot all of them, would likely be plotted on top of eachother in many areas. Would anyone have any suggestions on how I might go about downsampling the number of images shown so it appears a sensible density of images to look at when zoomed out, whilst showing progressively more as you zoom in? I was planning to plot them in pixel heights and widths so they retain the same size when zooming in/out. Datashade is obviously good at dynamically adjusting things but I’m not sure it would work in this instance. Thanks for any advice.

Edit. Having thought about it maybe a cds view with a Boolean that updates from a zoom callback in some way

In case anyone’s interested in this in the future, I managed to get it working with a cdsview, a callback based on lodEnd, and using datashade for doing the downsampling of which images to show. Works well even with around 15 000 small images with it only taking a couple of seconds to update when zooming.

I’m glad you found a pure-Bokeh solution. I still might encourage looking at Holoviews and their support for Bokeh+Datshader, since they have done alot of work towards robustness and efficiency. At least, I have run HV examples that work on billion+ point datasets that only take fractions of a second when panning/zooming, etc.