Primarily manually, though there could be other programmatic or scraping methods or something like feed43.com for example. This is one of the better versions of these sites in that it makes all of the richer data visible on the page. (I couldn’t seem to find one that provided data about which filter was used in the half dozen or so that I looked at.) It also makes it much easier to find the canonical URL at which the highest quality image lives, particularly with respect to the lengths that Instagram goes to hide this and other pieces of data.