devistuia
@devistuia
machine learning and remote sensing professor at EPFL. nature lover, diver and formerly musician.
ID:1132258856594223104
25-05-2019 12:15:22
640 Tweets
1,9K Followers
434 Following
Interested in addressing geographic biases, class imbalances, and presence-only data in species distribution modeling? Don't miss our poster at the Climate Change AI workshop at ICLR 2024. Collaborative work with Nina van Tiel Marc Rußwurm devistuia
We are waiting for your papers at the interface of vision 👁 and language 👄 for satellite 🛰 image analysis!
Deadline in September!
With Sylvain Lobry Nathan Jacobs BigEarth and J. Bazi!
Come join us at our spotlight 🤩 poster today at #ICLR2024 .
Konstantin Klemmer Robin Zbinden, and I will present it in Hall B #266 Wed 8 May 4:30 p.m. 6:30 p.m.
marcrusswurm.com/locationencode…
paper: openreview.net/forum?id=Puddu…
devistuia Esther Rolf
🙌⭐️ New paper by Gabeff Valentin! 🐻🌲
Vanilla CLIP (trained on natural images) is not precise when querying a camera trap dataset with text.
So, how can Vision-Language Models be adapted to wildlife images?
➡️ with WildCLIP: A specifically fine-tuned model. See paper ⬇️⬇️
WildCLIP is now published. The core goal is to ease the interaction between wildlife ecologists and their data using text. Here, we propose to adapt Vision Language Models to retrieve camera-trap events of interest. w/ devistuia A. Mathis Group Marc Rußwurm
link.springer.com/article/10.100…
Great work, the first of Gabeff Valentin PhD, which gives us a glimpse on what language models like CLIP can do to retrieve attributes and behaviour! Congrats to all the team! Alexander Mathis Marc Rußwurm
Excellent work by Thien Anh Nguyen unveils the patterns of migration of the upper tree line in Swiss mountains 🏔.
And with knowledge-guided deep learning!
This is wonderful research and real impact led by our very own Jonathan Sauder, proud of this collaboration with the LGB lab of EPFL and released in Methods in Ecology and Evolution on the 5th anniversary of the Transnational Red Sea Center !
🪸 + 🤿 + 🤖
🚨🌍🌲 New Paper in the Remote Sensing of Environment journal by Thien Anh Nguyen!
Takeaway: In our warming climate, trees start growing at higher elevations.
Thien Anh Nguyen shows this using forest maps from historical aerial images in Valais, Switzerland, created by deep learning.
🌲 🌲 By how much do forests move upwards? We analysed aerial images since the 1940s and developed a knowledge guided Deep learning 🤖 model to map forests across time!
Wonderful work led by Thien Anh Nguyen and just published in Remote Sensing of Environment!
