Hi all
How can i upload a list of birds detected by sound?
I have integrated a program for detecting birds by sounds, and I would like to add the functionality of uploading the detected birds to the inaturalist platform to make a bird census.
Similar to
https://support.ebird.org/en/support/solutions/articles/48000907878-upload-spreadsheet-data-to-ebird
Thank you in advance
Jon
Hi @jon welcome to the forum.
iNat does have the ability to upload observations via csv. There are numerous threads on the forum about this that you can check out, examples:
https://forum.inaturalist.org/t/bulk-upload-of-observations-metadata-to-inat-from-excel/13292
https://forum.inaturalist.org/t/specification-for-csv-upload-file/42366
https://forum.inaturalist.org/t/import-via-csv-can-this-include-additional-observation-fields/13013
That said, iNat does have a ban on machine generated content. There’s not a lot of detail in your post, so I can’t be sure, but a program that creates IDs for recordings without human input for upload would likely qualify as machine-generated content:
https://forum.inaturalist.org/t/community-guidelines-update-re-machine-generated-content/29029
https://www.inaturalist.org/pages/community+guidelines
One other thought is that, if the observations are already added to eBird, adding them to iNat as well may not be needed.
I have a microphone recording sounds in a forest and I’m processing wav files to identify birds with BirdNet.
I have birdnet for desktop &&birdnet-pi for linux server and I’m also working with the birdnet code, I could upload all the bird identifications to inat.
It would be the best way to do a bird census in my area.
Yes, if you are just using birdnet to ID the sounds and not making IDs yourself, I believe this would violate iNat’s ban on machine generated content, so users shouldn’t post observations based on this type of process. Another platform might be a better fit for this type of data.
Understood. Thanks for the clarification.
It’s good to have quality, but on the other hand I think iNat can filter by Confidence, publishing the identifications with very high Confidence.
With a large volume of quality data, anyone can make better AI predictions.
It’s a suggestion.
Best regards
The basic idea behind iNaturalist is the interaction between humans and nature, so machine based identification is not the focus of the platform.
Yes, iNat is for people powered observations. That said, if BirdNet detects an interesting bird, it’s people powered if you clip out the audio for that song and post that to iNat along with an ID and people can pitch in to agree or disagree with an ID.
I use BirdNet on an audio recorder in my garden and while it shows me the big patterns going on with birds in our garden, it’s wrong with its IDs some of the time. Its complete bird list from my garden is very wrong as it contains a long tail of mid-identifications. If you wanted to, for example, upload a list of the birds from your garden from a month, you’d definitely want to be uploading clips of the audio so people could check the IDs, and any species that BirdNet only detects rarely should be regarded with skepticism.
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.