Train a neural network using iNaturalist photos database

There’s a reason why iNat makes sure that people don’t decide on “dangerous” plants an fungi based on cv ids, it is not a really good idea to implement it with programm, only people can answer such questions with safety.

4 Likes

Certainly, I would steer way clear of affirmatively saying something was ‘edible’. However, there might be ways to structure and market such an app to indicate increased levels of danger, perhaps indirectly. The fact that there are multiple plant/object ID apps out there which are capable of suggesting edible and poisonous IDs suggests that it is possible to do in a lawyer-approved manner. Definitely another good reason to talk to a lawyer.

1 Like

Interesting, what a shame. Thanks to everyone for their valuable help! I will focus my efforts on other matters because of the copyright and safety limitations.

1 Like

You can check out some previous (somewhat involved) discussions on what the implications of various licenses are
https://forum.inaturalist.org/t/can-someone-explain-licensing-to-me-please/2691/10
https://forum.inaturalist.org/t/a-case-for-changing-the-default-license-to-not-include-a-nc-clause/18690
depending on how deep you want to get into the weeds.

2 Likes

Slightly tangential perhaps, but the latest text-to-image model had a press release from OpenAI today, and it looks insane :
https://openai.com/dall-e-2/

Presumably trained on billions of images, copyright and otherwise…

2 Likes

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.