Starting June 7th 2023, a set of changes implemented as part of the European GUARDEN and MAMBO projects, were made available that reaches deep into Pl@ntNet systems. The goal is always to support more species for every region, let’s dig:
1. New unified species list
Taxonomy and plant species names is a complex issue. Botanic institutes around the world publish lists of species for their countries (also called taxonomic backbones or checklists). But they also need to update them regularly as new species need to be added or previous ones need to be updated (change of author, synonmy, transfer of genera, etc).
We struggled in the past with this issue and decided to move to a global backbone provided by Kew Royal Botanical Gardens. The adopted checklist is named Plants Of The World Online (POWO). The POWO backbone provides up to date names and more importantly tries to cover plants of every country, while providing additional information.
2. New floras for all countries
To improve plant identification, we only display relevant plant species based on the selected flora. Until now, we were not able to cover each specific flora due to the lack of standards and heavy manual work required.
With the migration to POWO, we are now able to use a public standard (provided by the TDWG) which provides geographical areas for every country and region. The World Geographical Scheme for Recording Plant Distributions (WGSRPD) coupled with the POWO checklist now allows us to provide country-specific flora to each user.
We thus offer a flora for every region of the world and now cover all continents. The spatial breakdown used allows us to propose several regions on each continent, depending on the types of flora observed and territorial boundaries.
All existing observations will be automatically aligned on the new floras in the coming days.
3. New identification AI
Because the migration impacts every part of Pl@ntNet, we also needed to update our machine learning model (also called artificial intelligence), the one who identifies plants based on pictures. The update brings better performance overall and uses all the data you contributed to until January 1st 2023. We are proud to make it available for free through our app and API. For those wanting some technical insight, we’ve followed the state of the art, and have migrated away from Convolutional Neural Networks to Vision Transformers.
This changeover took almost two years to complete, involving numerous manual steps to verify the consistency of data and services throughout the migration.
Let us know what you think on your social networks or by contacting us.
Greetings,
The Pl@ntNet team.
