Over 3,000 species added to the Computer Vision Model (v2.15)!
Because v2.14 was a bit delayed, there's a big jump of 3,079 new taxa (93,369 taxa up from 90,290) in v2.15 which we released today. This new model was trained on data exported on July 28, 2024.
Here's a graph of the models release schedule since early 2022 (segments extend from data export date to model release date) and how the number of species included in each model has increased over time.
The graph below shows model accuracy estimates using 1,000 random Research Grade observations in each group not seen during training time. The paired bars below compare average accuracy of model 2.14 with the new model 2.15. Each bar shows the accuracy from Computer Vision alone (dark green) and Computer Vision + Geo (green). Overall the average accuracy of 2.15 is 88.8% (statistically the same as 2.14 at 87.9% - as described here we probably expect ~2% variance all other things being equal among experiments).
Here is a sample of new species added to v2.15:
Comentarios
Awesome! A lot of spiders made it :)
When I use identify mode and the observation has no taxon name (i.e. unknown) the suggestion tab has crazy suggestions. Like suggesting birds or zebras for an obvious plant.
Awesome!!
@andrewgillespie can you share an example?
The plants link won't load for me on my laptop or phone, even with strong internet—maybe too many taxa in the URL?
Also no plants - and I tried setting the filter to my obs - still ... loading. It is already location Cape Peninsula.
12 non-plant URLs.
Can we have more than the one plant URL?
3,079 new taxa - but - how many are plants?
@holyegg and @dianastuder - sorry about that - we split the Plant URL into 2
@andrewgillespie start the CV from a broad ID - it's a daisy or a $%#@ dicot.
Then use the CV as a tool (checklist, or visually similar), and delete your daisy ID.
Very cool! Excited to see more than a few of these in my local area.
Just a heads-up, it looks like the "Other Animalia" just links back to this blog post. I want to see if any new millipedes made it!
whoops - thanks @guerrichache, fixed
15 Chironomid taxa added. Only a handful of which are species. The lists provided do not take into account genera and other taxon levels which may have been learned.
One thing that I am seeing while looking at the heat maps of all of the species being added is just how many are in less observed areas like central america. It's really good to see more species being added!
@alexei_kouprianov we did it, Chilenocaecilius ornatipennis is now in the training model! :D
Very cool, I always love to have a look which arachnida are added! One question - what happens if a species that barely made it to the CV loses observations (e.g. correction or deleting).. will it stay in the new model or will it drop out again?
the species will drop out of the model - they usually do so based on taxon changes (e.g. a species split into many smaller ones) but sometimes from corrections/deleting
Thanks for the fast reply and clarification
Agregar un comentario