The posting of animal and plant photography to your Instagram, Snapchat, Rabadaba and Facebook social networks just got a lot easier. With a new app launched by iNaturalist, you can now identify flora and fauna, while you are getting up close and personal with them with your smartphone camera. This artificial intelligence [AI] software is triggered by a massive amount of big data which currently identifies approximately tens of thousands of species.
Naturalist Online Community
iNaturalist is a thriving online community guided and updated daily by an army of smartphone-toting naturalists, zoologists, biologists and and hobbyists. Professionals and amateurs alike can now take photos of the animals and plants they encounter, whether they uncover them in urban environs or in the wild.
“We started iNaturalist to explore how technology could be used to connect people to nature,” says Scott Loarie, one of the founders of the community. “But we also found that the observations and identifications people were sharing were an exciting new source of data for science and conservation.”
The stats are exceedingly impressive after it was launched six short years ago. Currently at 400,000 registered users, who have clocked up 4.8 million verifiable observations across 100,000 species, it’s moving at a fast clip to build its reservoir of identifiable lifeforms.
Artificial Intelligence only as good as . . .
It’s important to note however, AI is only as intelligent as the input it receives to train it. The iNaturalist version released on June 29 is surprisingly good according to several recent reviews.
“It has learned to recognize several species from unusual angles—like the head-on slaty skimmer dragonfly that I asked it to identify. It can even cope with species that come in various patterns,” noted Alex Shepard, one of Loarie’s colleagues.
“Asian ladybirds come with a lot of different characteristics—you might see one that’s mostly black with red spots, and another that’s red with black spots. But even the early versions of our system could understand that,” added Shepard.
Suggestions versus Bonafide IDs
Loarie’s team has tried to circumvent the risk of identifying in absolute terms, by designing their app to be almost self-conscious about its own limitations. Rather than providing firm identifications, it instead gives “suggestions” or “recommendations.”
For each photo you take, it offers ten possible species. To date, the app has been coming up with correct suggestions 78 percent of the time. It also gives one overarching recommendation, which varies in detail depending on how confident it is, based on its current accumulation of photos in the iNaturalist system.
So, iNaturalist isn’t quite a biological version of SoundHound—the app that identifies songs. It’s more a ‘work in progress,' which offers increasingly accurate suggestions depending on the information users provide. “We want something that’s always accurate even if it’s not precise,” says Loarie.
The iNaturalis community is very active. The founders and their developers continue to review the app’s ongoing responses. For instance, when a user recently identified a butterfly a ‘swallowtail,’ the community was quick to confirm it was specifically an “eastern tiger swallowtail [Papilio glaucus]. Good catch iNaturalist!