iNaturalist computer vision for video game

Hi
I’m an (very) novice developer hoping to use iNat’s computer vision for my video game. Is there any way for me to use a javascript request to send iNat a photo of a bird and get its identification?

2 Likes

welcome. the short answer is no. however, you can always train your own computer vision AI.

when you say you’re developing a video game, i can’t envision why you would even need to incorporate iNaturalist’s computer vision into something like that. if you’re making something like iNat’s Seek, then that’s not really a game. if you’re making something like a gamified version of iNaturalist, then you might need to be careful about how you approach this to make sure you’re not violating iNat terms of service.

5 Likes

Can you share more details? I’m having trouble thinking of a use case where accessing iNat’s computer vision would be more efficient than some other workaround.

1 Like

Thanks for the replies!

I’m making a game that plays webcams in Ecuador for a friend who runs ecolodges there. The goal for the player is to identify different species to earn points (sorta like wildlife bingo), but to check their identifications (my goal is to add players’ sightings as eBird checklists) I need species recognition software.

Training my own AI sounds like a fun project: what’s the easiest way to go about it?

1 Like

This research group has some papers on the subject. https://vision.cornell.edu/se3/projects/visipedia/

This is a simpler approach (that could potentially be extrapolated to your bird species) that would be easier for a first machine learning project. https://towardsdatascience.com/image-classifier-cats-vs-dogs-with-convolutional-neural-networks-cnns-and-google-colabs-4e9af21ae7a8

1 Like

i guess “easy” is in the eye of the beholder, but you might be interested in some things i mentioned here: https://forum.inaturalist.org/t/species-or-color-search-on-photos/17611/4. this thread may also be of interest: https://forum.inaturalist.org/t/how-to-download-taxa/3542. (in case it’s not clear, that latter thread references TensorFlow.) if you search the forum for computer vision training or other related keywords, you may also find other discussions that might be relevant.

3 Likes

No idea if it is usefull but Naturalis had an API in which one could do 10 identifications a day for free https://www.programmableweb.com/news/101-biology-apis-biocatalogue-biogrid-and-synergizer/2013/10/22

I just can not find the link, API

r. In samenwerking met
Waarneming.nl en Xeno-canto beheren en
publiceren we digitaal naast 42 mln. objecten en
soortbeschrijvingen ook 20 mln. waarnemingen. We
hebben in 2018 geïnvesteerd in genetische identificatie, scanningtechnieken en artificiële intelligentie. We hebben op basis van onze collectiesoftware kunnen trainen om automatisch soorten
te herkennen. We gebruiken die zelf om in de
collectie soorten op naam te brengen en hij is ook
beschikbaar als webservice die bijna 9.000 in
Nederland voorkomende plan

==
https://identify.biodiversityanalysis.nl/beta/observation/demo

https://biodiversityanalysis.nl/

1 Like

o’n API is er, specifiek op het herkenningsmodel dat wordt gebruikt in de app ObsIdentify. De informatie hierover kun je vinden op https://biodiversityanalysis.nl/ en https://biodiversityanalysis.nl/api (doorklikken naar PDF). Er zitten wel enkele restricties op die API. https://biodiversityanalysis.nl/files/User%20manual%20Nature%20Identification%20API%20-%20Public.pdf

==
https://identify.biodiversityanalysis.nl/beta/observation/demo

https://biodiversityanalysis.nl/

It is gone Een nieuwe ontwikkeling is het herkennen van foto`s via een web interface.
Deze web interface werkt via de servers van naturalis :
https://identify.biodiversityanalysis.nl/beta/observation/demo
Voordeel hiervan is dat je niet hoeft te zoeken naar de juiste app
Nadeel is dat dit wel dataverkeer kost en maximaal 10 herkenningen per dag per ip-adres mogelijk zijn.

==
https://identify.biodiversityanalysis.nl/beta/observation/demo

https://biodiversityanalysis.nl/