Hi heređź‘‹
I’m completely new to roboflow, and new to model training
and that why my question will maybe be too simple but it’s another way to thanks James about his blog’post…
In context, I’m implementing a bluesky bot which is able to identify plant via plantnet API
and now I’m studying the way to do the same thing but for birds !
cf. Identify Birds ! · Issue #57 · boly38/botEnSky · GitHub
So the basic idea is to find a free api to share happynness getting a post contribution augmented by bird identification result.
I quickly arrive on James blog post which explain very well how to use roboflow to do that
How to Use the Roboflow Bird Detection API (really amazing post )
A basic test is quickly done, and I continue with 14 random Internet bird pics, but unfortunately
it demonstrate some wrong identifications (3 ok of 14 sample). I thinks that’s because the success is limited to the current dataset knowledge.
the model only detects the species of around 30 birds,
So here is my question:
- is it possible to get a “bird v2” like dataset+training model supported by a community
that can improve over the year with open community contribution for example ? (in the same idea of wikiedia, open street map, open food fact…)
I think that maintaining an training a model is an huge and paining work, and that why I’m curios about an open contribution model to do that.
maybe roboflow is already able to do that I dont know ? what do you think ?
Thantks for your answer.