@joncounts agree re the humans, though the interaction fields are so rarely used! I have seen informal advice on iNat to actually repost the same image with two different taxa, to document interactions. Even projects that focus on interactions don't seem to use interaction fields? […]

@wragge yes, or the model gets confused! Gotta remove those bee pixels. Been having much strange fun with BioCLIP

@joncounts thanks for sending, very interesting. I started this to test out some expansive claims in papers like Pollock et al (2025) (https://doi.org/10.1038/s44358-025-00022-3), about the impact of AI on biodiversity knowledge. They argue for a move from piles of occurrences to a more […]

@wragge whoops, just made it public, should be good now!

Code, data and documentation of the pipeline is available here: https://github.com/mtchl/mining-bee-plant-interactions - if it's of use to others I would love to hear about it.

New work: Anthophiles uses machine learning to ID plants present in bee observations. It's an experiment in joined-up biodiversity data. https://anthophiles.mtchl.net/ #biodiversity #citizenscience #pollinators #insects

And an Australasian slimwing. Find me on iNat here if that’s your thing https://www.inaturalist.org/people/mitchellwhitelaw

#darwin #insects #macrophotography

More Rapid Creek dragonflies - a painted grasshawk, very happy perching on a roadside fence #darwin #insects #macrophotography

I’m in Darwin this week, meeting new people and critters on Larrakia country. This is a yellow-striped flutterer (Rhyothemis variegata), near Rapid Creek

#insects #macrophotography

Talk coming up, 16th Sept in Darwin (and online) "Sensing an inland wetland: glitches, patches and mosaics" - on the making of Mosaic. Register here […]

[Original post on post.lurk.org]