Is the CV trained on captive observations?

Is the Computer Vision at least partly trained on Captive Observations with photographs, location and date?

1 Like

yes it is

3 Likes

Good to know. I have been correcting mistaken IDs in certain captive observations, but I was wondering if this affected the CV (since there are a decent amount of misidentifications in captive observations of the species I identify)

Off-topic, but good grief you responded quickly (like 5 seconds after the topic was posted!)

6 Likes

:racing_car:___________

11 Likes

I believe they also need a community taxon.

2 Likes

See https://help.inaturalist.org/en/support/solutions/articles/151000170368-which-taxa-are-included-in-the-computer-vision-suggestions-

2 Likes

Oh dear! I do many plant identifications in south east asia. Malaysia, Thailand, and Indonesia come with lots of potted / garden plants not marked as cultivated - far too many. When I see such an observation, I just hit x - “cultivated”, and that’s it. Even if I see that it has a currently wrong id, I do not care to disagree.
That means that I now have to change that: disagree also on wrongly identified potted plants…

4 Likes

Correcting them is always a good idea. And I hope the usage of potted plants etc. for training stays that way. Since there is a very good chance of them to be identified correctly (after all, people know what they have planted). this might improve CV particularly for young plants.

2 Likes

If only I could share your confidence.

2 Likes