Jan Pauls
@jan_pauls_
Followers
75
Following
66
Media
5
Statuses
14
PhD Student @ University of Münster: Deep Learning & Remote Sensing
Joined May 2024
Our #AI4Science paper "Capturing Temporal Dynamics in Canopy Tree Height" is accepted at #ICML2025! We present the first 10m temporal map for Europe (2019-22), improving tall tree estimation & tracking change. Blog: https://t.co/SHfsSfLyvn Paper: https://t.co/5wWGbHM77c 🧵 1/7
1
2
6
Now with a blogpost (and the first one on my website!): https://t.co/YTcVkJTQte
Our paper "Estimating Canopy Height at Scale" has been accepted to #ICML24, where we significantly advance global canopy height mapping. w/ @maxzimmerberlin, U. Kelly, M. Schwartz, S. Saatchi, @ciais_philippe , @spokutta , @matin_brandt, F. Gieseke https://t.co/5vn882Zvtf 🧵 1/5
1
3
7
Please check our new pre-print with a Foundational Model base on contrastive learning that can map tree height, land cover, @GEDI_Knights LiDAR waveforms using multimodal satellite data https://t.co/qWhi4EIgZC
1
3
30
We are looking for a motivated PhD candidate @lsce @inrae @JWigneron with interest in deep learning and foundational models for mapping forest attributes in African dense forests using satellite data Please join your team ! Link and how to apply ⤵️ https://t.co/SrKdeYteA6
1
19
33
arXiv -> alphaXiv Students at Stanford have built alphaXiv, an open discussion forum for arXiv papers. @askalphaxiv You can post questions and comments directly on top of any arXiv paper by changing arXiv to alphaXiv in any URL!
131
2K
7K
A good time to share our #ICLR2023 paper: How I Learned to Stop Worrying and Love Retraining We explore sparsity-adaptive LR schedules and show that with proper LR care, simple pruning can outperform complex methods that 'learn' the sparsity. 📜 https://t.co/okFdRQcPGW 🧵1/n
I'm not really an expert on sparsity, but I enjoy using this template, and reminding about learning-rate, whenever I can. So I will:
1
9
38
Happy to present our work at #ICML24 in Vienna with @jan_pauls_! Join us at poster #412 on Tuesday from 11:30-13:00. See you there! https://t.co/Le5RDpDL12
Our paper "Estimating Canopy Height at Scale" has been accepted to #ICML24, where we significantly advance global canopy height mapping. w/ @maxzimmerberlin, U. Kelly, M. Schwartz, S. Saatchi, @ciais_philippe , @spokutta , @matin_brandt, F. Gieseke https://t.co/5vn882Zvtf 🧵 1/5
0
2
5
🌟 Join our Team in Berlin 🌟 We are seeking highly motivated PhD students to work on (efficient) Deep Learning, preferably with strong math/CS background and PyTorch experience. Happy to answer questions here, via DM or at #icml2024! Apply at https://t.co/8N0nJOIPlc! Please RT
5
7
14
Resources for further information: arXiv: https://t.co/5vn882Zvtf Online Viewer: https://t.co/gOeFAD43Xk GitHub: https://t.co/AvaNKb0g9S
0
3
17
Comparing our map with other global maps shows a noticeable improvement in resolution and vegetation structure detection. Additionally, our performance is almost on par with a regional map nearly bridging the quality gap between regional and global maps. 🧵 4/5
1
1
6
We introduced a pipeline to generate high-resolution canopy height maps using Sentinel-1/2 satellite images in combination with GEDI height measurements. Additionally, we employ a novel loss function reducing the impact of GEDI's systematic geolocation inaccuracy. 🧵 3/5
1
0
6
There is a significant disparity in quality between regional and global canopy height products in terms of visual quality and error metrics. This seems counterintuitive in the era of big models benefiting from large and diverse datasets. 🧵 2/5
1
1
5
Our paper "Estimating Canopy Height at Scale" has been accepted to #ICML24, where we significantly advance global canopy height mapping. w/ @maxzimmerberlin, U. Kelly, M. Schwartz, S. Saatchi, @ciais_philippe , @spokutta , @matin_brandt, F. Gieseke https://t.co/5vn882Zvtf 🧵 1/5
5
35
141