About

I’m a senior machine learning researcher at RISE Research Institutes of Sweden within the deep learning research group, where my main research interest is to develop ML methods for a broad range of environmental applications (e.g. climate adaptation and humanitarian aid causes). I co-host RISE Learning Machines Seminars, where Olof Mogren is the main host. I’m also affiliated with the Swedish Centre for Impacts of Climate Extremes (climes).

Before joining RISE, I conducted my doctoral studies in computer vision at the Faculty of Engineering, Lund University, under the supervision of Prof. Cristian Sminchisescu. In my research, I applied deep reinforcement learning to computer vision tasks such as object detection and active vision, with the aim of making visual recognition systems more flexible and adaptive. I successfully defended my PhD thesis Reinforcement Learning for Active Visual Perception in June, 2021.

Prior to my doctoral studies, I received a bachelor and master degree within engineering mathematics (in Swedish: Civilingenjör i teknisk matematik) from the Faculty of Engineering, Lund University. During this time, I spent a summer (2014) as a research intern at California Institute of Technology, where I improved an algorithm for clustering, a fundamental problem in machine learning and statistics. I also strengthened some theoretical recovery guarantees. This work resulted in a journal publication in JMLR. My supervisors were Assoc. Prof. Brendan Ames and Prof. Joel Tropp.

I’m an advocate for facing our difficult future (and present, for many), collectively and individually, so that we can try to reduce harm and suffering. I may write more about this under my personal page at some point, but for anyone interested in learning more about what is difficult about our future, I warmly recommend the podcasts The Great Simplification, Breaking Down: Collapse (and the follow-up podcast Building Up: Resilience), Planet: Critical, and Entangled World. Two other short and self-contained podcasts that I also recommend are Power: Limits and Prospects for Human Survival and Tipping Point: The True Story of “The Limits to Growth”. I also try to summarize some reasons why in this video.

Publications and preprints

GOMAA-Geo: GOal Modality Agnostic Active Geo-localization (NeurIPS 2024, to appear)Code

Flexible SE(2) graph neural networks with applications to PDE surrogates (arXiv 2024)Code & visualizations

Impacts of Color and Texture Distortions on Earth Observation Data in Deep Learning (2nd ML-for-RS Workshop at ICLR 2024)

NBS Initiative Position Paper: Embracing Nature-Based Solutions for Sustainable Development (2024) – Presented at the ECTP conference 2024

Creating and Leveraging a Synthetic Dataset of Cloud Optical Thickness Measures for Cloud Detection in MSI (Remote Sensing 2024)arXivCode & Data – Also presented (oral) at the 2nd ML-for-RS Workshop at ICLR 2024 – Fun fact: ML models developed in this project have been orbiting in space, via a collaboration with Unibap and D-orbit

Fully Convolutional Networks for Dense Water Flow Intensity Prediction in Swedish Catchment Areas (SAIS 2023)arXivCodeVideoPopular summary

Aerial View Localization with Reinforcement Learning: Towards Emulating Search-and-Rescue (1st ML-for-RS Workshop at ICLR 2023)SAIS 2023 paperML4RS workshop paperCodeVideo

Few-shot Bioacoustic Event Detection using a Prototypical Network Ensemble with Adaptive Embedding Functions (DCASE Workshop 2022)CodePopular summary

Embodied Learning for Lifelong Visual Perception (arXiv 2021)

Generating Scenarios with Diverse Pedestrian Behaviors for Autonomous Vehicle Testing (CoRL 2021)CodeVideo

Reinforcement Learning for Active Visual Perception (PhD thesis 2021)Video

Embodied Visual Active Learning for Semantic Segmentation (AAAI 2021)arXivVideo

Deep Reinforcement Learning for Active Human Pose Estimation (AAAI 2020)arXivCodeVideo 1Video 2Video 3

Semantic Synthesis of Pedestrian Locomotion (ACCV 2020)CodeVideoSpotlight video

Domes to Drones: Self-supervised Active Triangulation for 3d Human Pose Reconstruction (NeurIPS 2019)CodeVideo

Exact Clustering of Weighted Graphs via Semidefinite Programming (JMLR 2019)arXiv

Deep Reinforcement Learning of Region Proposal Networks for Object Detection (CVPR 2018)CodeVideo

Reinforcement Learning for Visual Object Detection (CVPR 2016)

Stray Light Compensation in Optical Systems (master thesis 2015)

Media and public communication

November 2024 (upcoming): Will participate in a panel discussion titled Is technology our only salvation from climate collapse? at Internetdagarna 2024.

October 2024 (upcoming): Guest lecture for a class of master students at Lund University on the theme AI for the environment.

October 2024: Invited presentation for IQ Samhällsbyggnad about AI and its many applications within the built environment sector.

August 2024: Participated in a panel discussion titled Science communication and AI – threats, challenges and opportunities, at the World Water Week in Stockholm.

August 2024: Invited presentation for Sally Dobberman on the topic AI considerations under post-growth scenarios.

January 2024: Commentary on AI for tackling climate change (Naturvetarna, in Swedish).

December 2023: Guest lecture for a class of master students at Lund University on the theme AI for the environment. Similar presentation also for Digitaliseringskonsulterna.

October 2023: Presentation for Lantbrukarnas Riksförbund (Federation of Swedish Farmers) about AI within the agricultural sector. Title: AI within horticulture: Where we are today, and where we may be headed.

October 2023: Presentation for a group of industry leaders within the industry sector in Skåne. Title: AI in practice: How to succeed with your first AI project.

September 2023: Commentary on AI for tackling climate change (Miljö och utveckling, in Swedish)

April 2023: Commentary on AI risk (Forskning och framsteg, in Swedish)

September 2022: AI for climate (Naturvetarna podcast, in Swedish)Extended conversation

May 2022: Keynote presentation about AI for tackling the climate challenges at Stockholm Tech.

January 2022: AI and climate change (RISE Learning Machines)Article about the talk

April 2021: Policy-based RL and applications within computer vision (RISE Learning Machines)

Past and ongoing projects

Aug 2024 - Aug 2027 (ongoing): AI-TOMO: Accelerated materials characterisation by AI and X-ray tomography. Funded by Vinnova.

May 2024 - May 2026 (ongoing): AI-based Power Production Models for Increased Wind Farm Efficiency together with PhD candidate Maria Bånkestad (among others). Funded by the Swedish Energy Agency.

2024 (ongoing): ML-based grazing detection in satellite imagery. Funded by the Swedish National Space Agency.

2023 - 2024 (ongoing): Involved in Agrifood TEF (see also RISE:s page for the project), where I’m exploring datasets and machine learning models for weed and crop detection.

Fall 2023 - Summer 2024: Towards efficient computational fluid dynamics simulations with physics-informed machine learning together with PhD candidate Maria Bånkestad (among others). Funded by Vinnova. Code and a preprint of our work is available here.

2023: ML for cloud optical thickness estimation together with the Swedish Meteorological and Hydrological Institute (SMHI), AI Sweden, Luleå University of Technology and the Swedish Forestry Agency. The work was published in the journal Remote Sensing (2024), got accepted as an oral at the 2nd ML-for-RS Workshop at ICLR 2024, and was presented as a poster at EUMETSAT 2023. Funded by Vinnova. One of the project deliverables was a developer event (Hackathon) for university students. Code & data are available!. Fun fact: ML models developed in this project have been orbiting in space, via a collaboration with Unibap and D-orbit.

Fall 2022: Pre-study about using ML for wetland monitoring in Sweden together with and funded by the Swedish Environmental Protection Agency. Code is available here.

2022: Industry project with Sightic Analytics: ML for detecting whether a human subject is under the influence of drugs. Funded by Vinnova.

Past and ongoing supervision

Below is a description of past and ongoing doctoral student supervision. Please see this page for past and ongoing master thesis supervision.

Summer 2023 (ongoing): Co-supervisor for PhD candidate Maria Bånkestad (RISE, Uppsala University). Academic supervisor: Prof. Thomas Schön (Uppsala University). Expected graduation: Fall 2024. Research topic: ML for the sciences (e.g. chemistry and physics).Until summer 2023, Erik Ylipää was Maria’s co-supervisor; I took over after Erik left RISE.

Honors and awards

March 2024: Top 30 reviewer of the ISPRS Journal of Photogrammetry and Remote Sensing for 2023.

June 2018: Outstanding Reviewer for CVPR 2018. Obtained an outstanding reviewer award for my CVPR 2018 review work.

March 2016: Teaching Assistant of the Year. Announced teaching assistant of the year 2016 for the chemistry students (K-sektionen) at the Faculty of Engineering, Lund University, for my contributions in the basic courses in calculus and linear algebra.

April 2014: SURF Scholarship at California Institute of Technology. Awarded a SURF scholarship from Caltech to work with Assoc. Prof. Brendan Ames and Prof. Joel Tropp during the summer of 2014.

Review work

Machine learning & computer vision. I regularly review top-tier ML/CV conference and journal papers (NeurIPS 2017, 2024; ICLR 2025; CVPR 2017 - 2019; ICCV 2019; ECCV 2018; AAAI 2018, 2021; BMVC 2022; CoRL 2022 - 2024; IJCV 2019 - 2021).

Remote sensing, environment & AI. I’ve also served as a reviewer for the ISPRS Journal of Photogrammetry and Remote Sensing (2023), Remote Sensing (2023), IEEE Transactions on Geoscience and Remote Sensing (2x 2024), and the International Degrowth Conference (2024).

Grant proposals. In addition to reviewing papers, I’ve also reviewed grant proposals for Climate Change AI (2024).