Summary¶
As a postdoctoral researcher at the EcoVision lab at University of Zurich, I collaborate with Jan Dirk Wegner to design deep learning methods for remote sensing and environmental applications.
Before joining UZH, I completed my PhD on "Efficient Learning on Large-Scale 3D Point Clouds" at IGN and ENGIE Lab CRIGEN, under the supervision of Loรฏc Landrieu and Bruno Vallet.
You like point clouds โ๏ธ ? You like trees ๐ณ ? You like satellites ๐ฐ๏ธ ? You like me ๐ค
๐ Publications¶
๐ผ Poster | ๐ค OralPaper | Webpage | Code
Elena Plekhanova, Damien Robert, Johannes Dollinger, Emilia Arens, Philipp Brun, Jan Dirk Wegner, Niklaus Zimmermann,
CVPR EarthVision workshop 2025 Poster ๐ผ
This work proposes a spatio-temporal sampling recipe for building pretraining sets for geospatial foundation models targeted for downstream macroecological applications. Simply put, we sample geolocations uniformly across the landmass and sample dates following local, phenology-informed seasons. We create a dataset SSL4Eco following this sampling recipe, on which we pretrain a model following the existing Seasonal Contrast approach. We show that our simple dataset construction recipe significantly impacts the performance on diverse downstream macroecological tasks, compared to other datasets, and pretrained models.
Paper | Webpage | Code
Johannes Dollinger, Damien Robert, Elena Plekhanova, Lukas Drees, Jan Dirk Wegner,
ICLR Tackling Climate Change with Machine Learning workshop 2025 Poster ๐ผ
Climplicit proposes to learn global, multi-temporal, implicit climatic representations to facilitate the use of climatic data ๐ฆ for downstream applications with minimal compute, memory, and deep learning expertise requirements ๐ฆ.
Paper | Webpage | Code
Kaan Karaman, Yuchang Jiang, Damien Robert, Vivien Sainte Fare Garnot, Maria Joรฃo Santos, Jan Dirk Wegner
ISPRS Geospatial Week 2025 Oral ๐ค
GSR4B proposes a method for producing high-resolution above-ground biomass ๐ชต๐ณ maps using guided super-resolution. More specifically, we show that it is possible to upsample 100m-resolution AGB maps to 10m resolution with the help of Sentinel-1/2 imagery ๐ฐ. The resulting estimates are more accurate than direct regression solely from Sentinel imagery and less prone to over-estimating low values and under-estimation high values.
Paper
Carlos Gomes, Isabelle Wittmann, Damien Robert, Johannes Jakubik, Tim Reichelt, Michele Martone, Stefano Maurogiovanni, Rikard Vinge, Jonas Hurst, Erik Scheurer, Rocco Sedona, Thomas Brunschwiler, Stefan Kesselheim, Matej Batic, Philip Stier, Jan Dirk Wegner, Gabriele Cavallaro, Edzer Pebesma, Michael Marszalek, Miguel A Belenguer-Plomer, Kennedy Adriko, Paolo Fraccaro, Romeo Kienzler, Rania Briq, Sabrina Benassou, Michele Lazzarini, Conrad M Albrecht
IEEE Geoscience and Remote Sensing Magazine 2025
This paper reviews the state of lossy neural compression ๐ค๐๏ธ, with a focus on its application and potential for the manipulation and analysis of Earth Observation data ๐ such as satellite imagery ๐ฐ and climate model predictions ๐ฆ.
Damien Robert, Hugo Raguet, Loรฏc Landrieu
Paper | Webpage | Code
3DV 2024 Oral ๐ค (top 5.3% submissions)
SuperCluster proposes a framework for efficient panoptic segmentation of large-scale
point clouds. We formulate the panoptic task as a graph clustering problem. We train
a model to predict desirable node and edge attributes to be used as input for a
downstream graph clustering algorithm. This allows for training a model with only
local supervision, without the need for non-maximum suppression, instance matching,
and without any prerequisite on the number of objects in the scene.
SuperCluster achieves new SOTA panoptic segmentation on indoor datasets S3DIS Area 5
(50.1 PQ (+7.8)) and ScanNetV2 (58.7 PQ (+25.2)), as well as outdoor datasets
KITTI-360 (48.3 PQ) and DALES (61.2 PQ).
๐ฆ 210k param. | โก Train S3DIS F5 in 4h | ๐พ 20M-point inference on 1 GPU
Damien Robert, Hugo Raguet, Loรฏc Landrieu
Paper | Webpage | Code
ICCV 2023 Poster ๐ผ (top 26.8% submissions)
SPT is a superpoint-based transformer ๐ค architecture that efficiently โก
performs semantic segmentation on large-scale 3D scenes. This method includes a fast
algorithm that partitions ๐งฉ point clouds into a hierarchical superpoint structure,
as well as a self-attention mechanism to exploit the relationships between
superpoints at multiple scales.
We reach SOTA on S3DIS 6-Fold (76.0 mIoU),
KITTI-360 Val (63.5 mIoU), and DALES
(79.6 mIoU)n with:
๐ฆ 212k param. | โก Train S3DIS F5 in 3h | โ SPG preprocessing รท7
Damien Robert, Bruno Vallet, Loรฏc Landrieu
Paper | Webpage | Code | Video
CVPR 2022 Oral ๐ค and Best Paper finalist ๐ (top 0.4% submissions)
An end-to-end multi-view aggregation method for 3D semantic segmentation from images and point clouds. We reach SOTA on S3DIS and KITTI-360 without requiring point cloud colorization, meshing, or depth sensors: just point clouds โ, images ๐ธ, and their poses.
๐ Short Resume¶
- 2024 - Now Postdoctoral researcher on deep learning for remote sensing and environment with Jan Dirk Wegner
- 2020 - 2024 PhD student on "Efficient Learning on Large-Scale 3D Point Clouds" supervised by Loรฏc Landrieu and Bruno Vallet
- 2022 International Computer Vision Summer School
- 2017 CNRS AI Fall School IAยฒ , Multi-disciplinary course for AI students and researchers
- 2011 - 2015 Master of Science at Ecole Centrale de Lyon
๐ Awards¶
- 2025 AFRIF 2024 PhD Award accessit
- 2024 ECCV 2024 outstanding reviewer
- 2022 DeepViewAgg shortlisted as best paper finalist at CVPR 2022
๐ผ๐ค Talks & Presentations¶
๐ผ Poster | ๐ค Oral- ๐ค 03/2024 Presenting our work SuperCluster at 3DV 2024 (Oral ๐)
- ๐ 02/2024 Starting a new position as a postdoctoral researcher in the EcoVision lab at University of Zurich ๐
- ๐ค 01/2024 PhD defense on Efficient Learning on Large-Scale 3D Point Clouds at IGN, in presence of Cรฉdric Demonceaux, Patrick Pรฉrez, Siyu Tang, Duygu Ceylan, Loรฏc Landrieu, and Bruno Vallet ๐
- ๐ 10/2023 Our work SuperCluster was accepted for an oral presentation at 3DV 2024 ๐
- ๐ค 10/2023 Presenting IGN's research on Large-Scale 2D and 3D Learning to the National Land Survey of Finland (NSL)
- ๐ค 10/2023 Presenting our work SuperCluster to the Ecole des Ponts, IMAGINE lab
- ๐ค 07/2023 Presenting our work Superpoint Transformer at ICCV 2023
- ๐ค 09/2023 Presenting our work Superpoint Transformer to the ETH Zรผrich, Computer Vision and Geometry lab
- ๐ค 09/2023 Presenting our work Superpoint Transformer to the ETH Zรผrich, Photogrammetry and Remote Sensing lab
- ๐ 07/2023 Our work Superpoint Transformer was accepted at ICCV 2023 ๐
- ๐ค 06/2023 Presenting our work Superpoint Transformer to the ENGIE Lab CRIGEN
- ๐ค 05/2023 Presenting our work Superpoint Transformer to the University of Zรผrich, EcoVision lab
- ๐ค 05/2023 Presenting our work Superpoint Transformer to the Samp R&D lab
- ๐ค 05/2023 Presenting our work Superpoint Transformer to the Valeo.ai lab
- ๐ค 12/2022 Presenting Self-Supervised Learning for Computer Vision to the LASTIG lab
- ๐ค 11/2022 Presenting IGN's research on Large-Scale 2D and 3D Learning to the German Federal Agency for Cartography and Geodesy (BKG)
- ๐ผ 07/2022 Presenting our work DeepViewAgg at ICVSS
- ๐ค 06/2022 Presenting our work DeepViewAgg at CVPR 2022 (Best Paper finalist ๐)
- ๐ค 06/2022 Had the honor to be interviewed for the CV News Best of CVPR issue
- ๐ค 06/2022 Presenting our work DeepViewAgg to the Ecole des Ponts, IMAGINE lab
- ๐ผ 06/2022 Presenting our work DeepViewAgg at ISPRS 2022
- ๐ค 05/2022 Presenting our work DeepViewAgg to the Ecole Polytechnique, LIX lab
- ๐ค 04/2022 Presenting our work DeepViewAgg at the AI4GEO project seminar
- ๐ผ 03/2022 Presenting our work DeepViewAgg at the IGN-ENSG Research Days
- ๐ค 01/2022 Presenting our work DeepViewAgg at the Information, Signal, Image and Vision research group seminar
- ๐ผ 05/2021 Presenting our work DeepViewAgg at the IGN-ENSG Research Days
๐ Teaching¶
- 05/2024 NeRFs, and Diffusion at University of Zurich (Course Instructor, M2, 5.5 hours)
- 01/2023 Deep Learning for Remote Sensing at ENSG (Course Instructor, M2, 13 hours)
- 06/2022 3D Deep Learning for Remote Sensing at ISPRS 2022 (Tutorial Instructor, 1 day)
- 05/2022 3D Deep Learning, Torch-Points3D & DeepViewAgg at ENGIE Lab CRIGEN (Tutorial Instructor, 1 day)
- 01/2022 Deep Learning for Remote Sensing at ENSG (Course Instructor, M2, 9 hours)
- 11/2020 Deep Learning for Computer Vision at Ecole Polytechnique (Teaching Assistant, M1, 12 hours)
๐ Affiliations¶
- EcoVision lab, University of Zurich, Zurich, Switzerland