A Very Niche Update

Great news for Ranunculus micranthus fans! iNat's newest computer vision (CV) model just dropped (check out their blog post for more information), the last update was in back in July 2021. In the previous model, this taxon did not have enough Research Grade images to be included in training data, so it was not a species that the CV model "knew" or was apt to suggest. As a result, the top CV suggestion for observations of R. micranthus and a few other small-flowered buttercup species were almost exclusively for the much more widespread and common R. abortivus, leading to many buttercups being misidentified, sometimes even erroneously reaching Research Grade. It took a lot of meticulous QC work from dedicated Ranunculus micranthus fusspots, but over time many of these misidentifications were corrected, and more new observations were identified correctly, resulting in a dataset that finally met inclusion criteria.

What this means for you

If you create an observation of R. micranthus, iNat's computer vision will now probably actually suggest the correct species! (Homework: give the CV a test drive if you've got any spare micranthus pics and report back)

The bigger picture

This obviously affects much more than those of us with Ranunculus micranthus vs. abortivus monomania, and if you read through iNat's blog post you can see how many more species this new model has been trained on, and you may even notice an improvement in CV suggestions for various other species as well. Thanks to everyone who works hard to correctly ID species, whether it's a buttercup or not -- your efforts help to improve computer vision on iNat and make it as bang-on accurate as it can be for all of us. :)

Posted on April 14, 2022 04:45 PM by vvoelker