I’m thrilled to announce that my paper with @MExpositoAlonso and @MeganRRuffley tracking rapid plant species changes with deep learning is live at PNAS!.

🎉🎉🎉.

Be sure to stop by and check out our poster at the afternoon session of the Tackling Climate Change with Machine Learning workshop at NeurIPS!! With the amazing @elenaXresearch 🤗

RT @MExpositoAlonso: Very excited to see our @PNASNews paper led by @leg2015 on deep learning maps of plant species using remote sensing….

One ocean, two planes, three trains, and 35 stickers later (long story), my first ECCV poster is finally on the books!

Hey #ECCV24 be sure to stop by tomorrow’s morning poster session to check out CRISP, our new multimodal contrastive pretraining strategy for natural world imagery task that the amazing @andyvhuynh and I have been working on over the past year! .

RT @MExpositoAlonso: Read our latest paper @PNASNews! . The first #DeepLearning project in our lab led by Lauren Gi….

And stay tuned for future updates on how @dneves_ecology and I are working on tools for this here in Brazil! 🌱

New automated biodiversity monitoring tools powered by deep learning and large-scale ecological datasets like Deepbiosphere will prove crucial to monitoring progress towards the United Nations’ Global Biodiversity Framework’s targets by 2030 worldwide

And most importantly, Deepbiosphere can detect temporal changes in biodiversity, including the lasting effects of severe megafires on plant communities in the western U.S.

When it comes to changes in biodiversity, Deepbiosphere can also detect spatial changes in biodiversity driven by habitat fragmentation and ecotones in ecologically diverse habitats

Beyond having excellent predictive performance, Deepbiosphere can efficiently detect species of all shapes and sizes, including both old-growth and clear-cut regrowth redwood groves, along with understory species with distinct preferences for different forest ages

Our model, Deepbiosphere, is a multimodal deep learning model that combines the pattern-detecting capability of CNNs with a simple climate prediction head to leverage both high-resolution remote sensing imagery and climate data simultaneously

To build a model that can predict thousands of plant species simultaneously at high resolution, we built a new machine learning dataset of over 650,000 aerial images paired to iNaturalist observations

Looking forward to presenting in 2 weeks at the Spatial Biology Workshop! Can’t wait to share my work on mapping biodiversity patterns in California with the broader imaging community & see how lessons learned from macroecology can translate to the microscopic level + vice versa!.

it’s not a bad problem to have though!.

There are just too many cool posters for me to see at #ESA2023 that I don’t think I can catch them all 😭.

Amazing scientists like these are why I’m here trying to become an ecologist, not just a computer scientist 😁.

Most of all, I want to give a huge thank you to our speakers today @KD_Chadwick @sarameghanbeery @Diego_EllisSoto Ruth Oliver, Mike Buchalski, and Andy Kleinhesselink, and special thanks to co-organizers @MeixiLin and Dave Schimel for all your amazing help!.

Had an amazing time moderating our inspire session on open questions facing biodiversity monitoring at scale! The discussion was especially salient with great audience questions ranging from data sovereignty to open data sharing practices, to mitigating the loss of the long tail.