Published:
Aleksis Pirinen has joined the core team of the newly formed GEO AI4EO working group, led by Prof. Yifang Ban (KTH Royal Institute of Technology). Group on Earth Observations (GEO) is the UN body for EO, and the AI4EO working group will play a crucial role in GEO’s post-2025 vision “Earth intelligence for all”.
Published:
The 2025 Nordic Workshop on AI for Climate Change will take place in Gothenburg, Sweden, May 13th 2025! The event is organized by us in the DL-group.
Published:
Climate AI Nordics is a network for researchers in the Nordics working on problems related to tackling climate change using AI and machine learning. We at the DL-group have now offially initiated this network and the website is online!
Published:
Two FORMAS grants obtained: (i) Grey to green: Using AI to detect and prioritize conversion of impervious surfaces to multifunctional nature-based solutions and (ii) PRACTICAL WISION – Automatically identifying and mapping different weed species through the practical application of AI-based image analysis models. The DL-group will lead a work package each within the respective projects.
Published:
The paper GOMAA-Geo: GOal Modality Agnostic Active Geo-localization has been accepted to NeurIPS 2024! Code is available here.
Published:
The paper From Weak to Strong Sound Event Labels using Adaptive Change-Point Detection and Active Learning by Martinsson et al. was nominated for best student paper at EUSIPCO 2024.
Published:
ML models for cloud optical thickness estimation developed my members of the DL group (available here) have been running on-device in satellites orbiting in space. Read more about it here.
Published:
The paper From Weak to Strong Sound Event Labels using Adaptive Change-Point Detection and Active Learning by Martinsson et al. was accepted at EUSIPCO 2024.
Published:
We are happy to announce that we have received funding from the Swedish Energy Agency for the project AI-based Power Production Models for Increased Wind Farm Efficiency. Project summary: Wind energy is a promising source of power but is not easy to utilize effectively. Wind farms consist of many turbines that have complex interactions with each other and their surroundings. Factors such as terrain, wind trail effect (wake) between turbines, and ice accumulation on the blades influence the amount of power generated. Predicting the power output of wind farms typically relies on time-consuming simulations, but an emerging paradigm based on AI can drastically speed up prediction methods while maintaining their reliability. In this project we will develop new methods that use AI trained on real-world data to get accurate prediction of wind farm power output at a low computational cost. As the turbines and their relationships can be seen as a graph, we will use Graph Neural Networks (GNNs) to model them. Our method can have a big impact as the number of wind farms keeps growing, improving their efficiency and planning, and enabling more sustainable and affordable energy.
Published:
The DL group will be represented at the Visipedia workshop hosted at the Pioneer Centre for AI in Copenhagen (to be held on April 12, 2024). The Visipedia project is jointly led by Serge Belongie’s group (University of Copenhagen) and Pietro Perona’s group (Caltech). Visipedia’s goal, broadly speaking, is to make computer vision systems that can be queried and used by large communities of experts to help foster the curation and generation of new knowledge.
Published:
Published:
The DL group will meet you at ICLR 2024! We have two papers (see here and here) to be presented at the ML4RS workshop, and Olof Mogren will be a panelist at the Tackling Climate Change with Machine Learning Workshop organized by Climate Change AI.
Published:
The NBS initiative position paper Embracing Nature-Based Solutions for Sustainable Development was presented at the ECTP conference 2024.
Published:
The paper Creating and Leveraging a Synthetic Dataset of Cloud Optical Thickness Measures for Cloud Detection in MSI was accepted for the journal Remote Sensing (2024). Code, models & data is available here.
Published:
We wecolme eight new master thesis workers to the DL group during the spring of 2024. The four associated master theses will revolve around diverse topics: ML for detecting coffee berry disease, data-efficient ML for EO, active learning for soundscape analysis, and distributed ML.
Published:
Published:
Published:
Three Vinnova grants obtained (“Emerging Technology Solutions”): Towards efficient computational fluid dynamics simulations with physics-informed machine learning, Active learning for ecological monitoring, and Structural causal models for distributional shift in federated learning.
Published:
The paper Fully Convolutional Networks for Dense Water Flow Intensity Prediction in Swedish Catchment Areas was accepted for SAIS 2023. Code & models are available here. A video presentation can be found here.
Published:
The paper Aerial View Localization with Reinforcement Learning: Towards Emulating Search-and-Rescue was accepted for the 1st ML-for-RS Workshop at ICLR 2023. Code is available here. A video presentation is found here.
Published:
Published:
Published:
Published:
Published:
Published: