This is good for now, I just wanted to note that it would be good to eventually have these scaling factors as part of the main icon finding pipeline (rather than just the benchmarking), so that we can adjust the size of the inputs more easily in case they do need to be scaled to the correct size range.
Originally posted by @ewadkins in #30