I thuoght iNat was goingto do something with sounds and read something about open code for European birds but that part was in Dutch. I just do not know if it is allowed to post this kind of information on this forum. I read the code was open (Arise is from Naturalis museum Leiden). I just post it may be some have iterest in reading it.
A short introduction to Machine Learning results
Door Bob Planqué
oktober 15, 2024
From day 1 in 2005, Xeno-canto has been about recognising sounds as much as about collecting and curating them. In the mean time, recognising sounds, especially automatically, has proven to be a lot tougher than collecting and curating them :-) . Interestingly though, the fact that the collecting and curating part of XC has gone very well has been key in getting automatic recognition of the ground. Well, that and the fact that we chose an open format, which has made the recordings here available for research groups worldwide.
Since 2014, important progress has been made during the yearly BirdCLEF competions for which Xeno-canto provided recordings for training and testing. The first Machine Learning (ML) algorithms appeared there in 2016. All of you must be aware of their success and their use in various recognition apps for phones that allow one to identify bird sound. We are very pleased with the role of XC in that regard.
Despite being involved from the get-go we’ve so far not managed to implement such algorithms in a meaningful way on XC. But here we are!! XC has started integrating ML algorithms for sound recognition into the platform by mid October 2024. It is the first cautious step on a trip in which we want to leverage automatic recognition on XC in various ways.
The ideas executed now have been around ever since we started, so it seems for us to have taken a very long time. In the mean time the environment in which XC operates has changed considerably, with very large initiatives being set up to develop and support automatic recognition. For us the “Arise” project in which our main partner Naturalis cooperates is key. We’ve stayed the course, we are convinced it is still early days for sound recognition, and XC are there for the ride. Let’s enjoy it :-)
Th
https://xeno-canto.org/article/299
https://guarden.org/public-deliverables/
https://guarden.org/wp-content/uploads/2024/02/D2.5-GUARDEN-system-architecture-design-and-technical-specifications-1.pdf