Outreach2024-03-25T15:43:07+01:00


Presentations

  • Segment Anything Model for Refugee-Dwelling Extraction with Few Samples from High-Resolution Satellite Imagery. Yunya Gao and Hui Zhao, 18th Conference on Location Based Services (LBS 2023), Ghent, Belgium, 20.11.2023
  • Segment Anything Makes It Easier to Produce Deepfake Satellite Imagery, Yunya Gao, International Data Week 2023, Salzburg, Austria, 23.10.2023.
  • Opening Plenary: Welcome to IDW 2023 ‘Spatial Data Science: Geographic Context Matters’. Getachew Workineh Gella, International Data Week 2023, Salzburg, Austria, 23.10.2023.
  • Dwellings as anomalies: integrating VAEs for unsupervised object counting in humanitarian emergency response. Getachew Workineh Gella, Symposium of the International Future Lab AI4EO-2023, Technical University of Munich, Munich, Germany, 09.10.2023.
  • Deep Learning Powered Non-Local Speckle Filtering of Sentinel-1 Imagery and Its Potential for Humanitarian Action. Getachew Workineh Gella, International Geosciences and Remote Sensing Symposium (IGARSS) 2023, Pasadena, California, USA, 18.07.2023.
  • Variational Autoencoders for Unsupervised Object Counting from VHR Imagery: Applications in Dwelling Extraction from Forcibly Displaced People Settlement Areas. Getachew Workineh Gella, International Geosciences and Remote Sensing Symposium (IGARSS) 2023, Pasadena, California, USA, 19.07.2023.
  • Integration of sparse multi-source earth observation data with deep learning for crop type and yield estimation in smallholder farming areas. Getachew Workineh Gella, 13th International Symposium on Digital Earth on Earth intelligence to understand and protect our Living Planet, Athens, Greece, 12.07.2023.
  • People in need – How satellite data reveal food and nutrition demand, Stefan Lang, CIVIS Global Days, University of Tübingen, Germany, Neue Aula, 25.05.2023.
  • Towards a geoprivacy-aware sharing of very high-resolution satellite imagery for humanitarian applications, Getachew Workineh Gella, IEEE GRSS second workshop on Remote Sensing Data Management Technologies in GeoScience 2022 (RSDM-GeoSci 2022), Paris, France, 02.12.2022.
  • Spatially Transferable Dwelling Extraction from Multi-Sensor Imagery in IDP/Refugee Settlements: A Meta-Learning Approach, Getachew Workineh Gella, International Symposium on Applied Geoinformatics 2022, Chania, Greece, 14.10.2022.
  • R & geospatial data: identifying vulnerable populations at risk, Linda Menk, GeOnG 2022, Chambéry, France, 26.10.2022.
  • Role of Emerging Technologies for Humanitarian Response: current advances and prospects of integrating AI with Earth observation?, Getachew Workineh Gella, GeOnG 2022, Chambéry, France, 26.10.2022.
  • Earth observation to foster peace building, Stefan Lang, ASPR 40 Years Anniversary Ceremony and Conference, Schlaining, Austria, 05.10.2022.
  • Why a session on Hybrid AI? Complementary research approaches in machine learning and the need for a common language to bridge between research domains and user domains, Stefan Lang, GI_Salzburg22, Salzburg, Austria, 06.07.2022.
  • Building extraction in the humanitarian context: constraints, challenges and opportunities for hybrid AI experiments from 10+ years of data collection, Yunya Gao and Getachew Workineh Gella, GI_Salzburg22, Salzburg, Austria, 06.07.2022.
  • Assessing the Influences of Band Selection and Pretrained Weights on Semantic-Segmentation-Based Refugee Dwelling Extraction from Satellite Imagery, Yunya Gao, AGILE 2022, Vilnius, Lithuania, 16.06.2022.
  • Explainable AI for Humanitarian Action, Stefan Lang and Dirk Tiede, International Conference on Image Processing Theory, Tools and Applications (IPTA), Salzburg, Austria, 22.04.2022.
  • Big data and humanitarian mapping, Stefan Lang, Afro-Asian Institute, public seminar, Salzburg, Austria, 20.1.2022.
  • Geospatial and EO-based technology for humanitarian action, Stefan Lang, MSF Open Board Meeting, 27.10.2021.
  • Mask R-CNN-Based Building Extraction from Satellite Data in Humanitarian Action, Dirk Tiede, ESRI User Conference 2021: GIScience session, online, 12.7.-15.7.2021.
  • Suggestions on the Selection of Satellite Imagery for Future Remote Sensing-based Humanitarian Applications, Yunya Gao, Symposium of the International Society for Digital Earth 2021 – Special Forum on Humanitarian Action, Salzburg, Austria, 08.07.2021.
  • Multi-feature Sample Database for enhancing Deep Learning Tasks in Operational Humanitarian Applications, Stefan Lang, Symposium of the International Society for Digital Earth 2021 – Special Forum on Humanitarian Action, Salzburg, Austria, 08.07.2021.
  • Testing Transferability of Deep Learning-based Dwelling Extraction in Refugee Camps, Gella Getachew Workineh, Symposium of the International Society for Digital Earth 2021 – Special Forum on Humanitarian Action, Salzburg, Austria, 08.07.2021.
  • Extraction of Dwellings of Displaced Persons from VHR Radar Imagery – Current Challenges, Future Perspectives, Andreas Braun, Symposium of the International Society for Digital Earth 2021 – Special Forum on Humanitarian Action, Salzburg, Austria, 08.07.2021.
  • Evaluation of DEMs Derived from Multi-Date Satellite Stereo Imagery for Urban Areas, Lorenz Wendt, Symposium of the International Society for Digital Earth 2021 – Special Forum on Humanitarian Action, Salzburg, Austria, 08.07.2021.
  • Determining mortality rate by counting graves in satellite images, Lorenz Wendt, Flavio Finger (MSF), MSF Scientific Day 2021, online, 10.06.2021.
  • Mask R-CNN based Mapping of Dwellings in Refugee Camps from Very High-Resolution Satellite Imagery, Getachew Workineh Gella, 40th EARSeL symposium 2021, Warsaw, Poland, 08.06.2021.
  • Mask R-CNN-Based Building Extraction from VHR Satellite Data in Operational Humanitarian Response – an Example Related to Covid-19 Response in Khartoum, Sudan, Dirk Tiede, Esri Imagery and Remote Sensing Educators Summit 2021, online, 29.04.2021.
  • Supporting urban sanitation management through the integration of EO-based indicators, Barbara Riedler, EARSeL joint workshop 2021: Earth Observation for sustainable cities and communities, online, 1.4.2021.
  • Flood mapping with radar, Andreas Braun, MSF GIS week 2020, 24.09.2020.
  • Radar imagery for humanitarian action, Andreas Braun, MSF GIS week 2020, 01.10.2020.



Posters

  • Sentinel-1 SAR Data For Building Damage Assessment After The Turkey-Syria Earthquake, February 2023. FRINGE 2023, Jaggy, N., Dabiri, Z. , Braun, A., Jessen, L., Lang, S., Nafieva, E., Leeds, UK, 11.-15.09.2023.
  • Unsupervised Domain Adaptation for semantic segmentation of dwellings with Unbalanced Optimal Transport. International Conference on Learning Representations (ICLR) 2023, Workshop on Machine Learning for Remote Sensing. Kigali, Rwanda, P. Sharma, N. Courty, G. Gella, S. Lang, May 2023. https://iclr.cc/virtual/2023/workshop/12829#wse-detail-14901.
  • Earth Observation Services for Humanitarian Action, Living Planet Symposium, Bonn, S. Lang, P. Füreder, L. Wendt, D. Tiede, A. Braun, B. Schernthanner-Hofer, B. Riedler, A. Alobaidi, K. Herzinger and E. Rogenhofer, 23.-27.05.2022.
  • Delineating and understanding urban structures through a multi-data approach, Living Planet Symposium, Bonn, B. Riedler, and S. Lang,  23.-27.05.2022.
  • Assessing risk of forcibly displaced persons: Creation of a digital elevation model of the island of Bhasan Char from Sentinel-1 Stripmap products. Living Planet Symposium, Bonn, A. Braun, 23.-27.05.2022.
  • Supporting humanitarian missions with ALOS PALSAR-2, Joint PI Meeting of JAXA Earth Observation Missions, online, A. Braun, January 2021.

Webinars and Workshops

  • EARSeL Joint virtual workshop: Earth Observation for sustainable cities and communities (30.03.2021 – 21.04.2021)
    Abstract:  Supporting urban sanitation management through the integration of EO-based indicators (Riedler B., Nödel J., Siber R., Andriessen N., Strande L., Lang S., 2021, in: Abstract book Earth Observation for sustainable cities and communities )
    Recording of presentation: for workshop participants

Events

GEOHUM was well presented at the International Data Week 2023 which took place in Salzburg from 23.-26. October 2023.

Gella Getachew was invited as speaker at the opening plenary, Yunya Gao gave a presentation on ‘Segment Anything to produce deepfake satellite imagery’ and we had a booth at the exhibition area.

Programme

Forum D20: “Hybrid AI Part 1: State of the Art”, GI_Studio, 6th July 2022, 9:00 – 10:15
  • Assoc. Prof. Stefan Lang, Head of Christian Doppler Laboratory GEOHUM, Paris- Lodron-University of Salzburg, Austria
    Why a session on Hybrid AI? Complementary research approaches in machine learning and the need for a common language to bridge between research domains and user domains Download presentation
  • Dr. Devendra Singh Dhami, Independent Research Group Leader and Post Doctoral Researcher, hessian.AI and AIML Lab, TU Darmstadt, Germany
    Hybrid AI: The Way Forward with Applications to Satellite Data Download presentation
  • Nicolas Longépé, Earth Observation Data Scientist ESA Φ-lab, Frascati, Italy
    R&D on hybrid AI in remote sensing at the ESA Φ-lab (online)
  • Dr. Omid Ghorbanzadeh, Postdoctoral Researcher in the AI4RS group at IARAI (Institute of Advanced Research in Artificial Intelligence), Vienna, Austria
    Multitemporal EO data analysis, coupling OBIA with DL models
Forum D22: “Hybrid AI Part 2: Defining a Data Challenge” GI_Studio, 6th July 2022, 11:00 – 12:15
  • Dr. Yonghao Xu, Postdoctoral Researcher in the AI4RS group at IARAI (Institute of Advanced Research in Artificial Intelligence), Vienna, Austria
    Universal Adversarial Examples in Remote Sensing: Methodology and Benchmark (online) Download presentation
  • Yunya Gao, Getachew Workineh Gella, Christian Doppler Laboratory GEOHUM, Paris- Lodron-University of Salzburg, Austria
    Building extraction in the humanitarian context: constraints, challenges and opportunities for hybrid AI experiments from 10+ years of data collection Download presentation
  • Open discussion on steps forward towards a data challenge on hybrid AI Download presentation

Graphic Story from Hybrid AI, GI_Salzburg22

Why Hybrid AI?

Convolutional neural networks (CNN) have become ubiquitous in remote sensing and close-range image segmentation and classification, but they need many training data, and the inner workings of deep learning approaches are often hidden from the user.

Rule-based classifiers and physical models and simulations on the other hand provide well-explainable results, but often lack adaptability. The synthesis of both data-driven and theory-driven approaches form a new generation of classifiers, sometimes called “Explainable Artificial Intelligence”, “Physics-Aware AI” or “Hybrid AI”.

Wait! What exactly is Hybrid AI?

So what are “Explainable Artificial Intelligence”, “Physics-Aware AI” and “Hybrid AI”? Or is Knowledge-based AI the better term? Many terms for the same thing? For similar things? Or related concepts? We need to clarify a bit what concepts are around, and how to make the best out of them for Earth observation applications.

Convolutional neural networks (CNN) have become ubiquitous in remote sensing and close-range image segmentation and classification, but they need many training data, and the inner workings of deep learning approaches are often hidden from the user.

Rule-based classifiers and physical models and simulations on the other hand provide well-explainable results, but often lack adaptability. The synthesis of both data-driven and theory-driven approaches form a new generation of classifiers, sometimes called “Explainable Artificial Intelligence”, “Physics-Aware AI” or “Hybrid AI”.

What makes satellite or drone image analysis a special case? What specific challenges (timeliness, accuracy, reliability) do we encounter in humanitarian action or other specific domains?  What role do classical spatial concepts play in today’s automated image analysis? How can we infuse our prior knowledge of a scene, the observation conditions or the quality of the labels into a classifier to deal with poor samples or reduce the need of a multitude of samples? What are best practices? How can we structure our knowledge in a machine-readable and transferable way? Are ontologies the way to go? What are standards or best practices, like STAC, to describe samples? How far are these concepts from being widely used?

In this Special Forum, we want to develop a common understanding and share both practical and theoretical experience to answer these questions.

From thinking to action. Let’s challenge our colleagues (and ourselves)!

We aim to go one step further to draft a data challenge, in which we want to invite researchers from the wider scientific community to showcase their hybrid AI approach on a benchmarking dataset. What are interesting and challenging cases we should set up? How should we prepare the data to conduct and evaluate these challenges?

The ideal outcome of this Forum is a deeper insight into hybrid AI, and the team building to design, promote, conduct, and evaluate a data challenge.

Invitation to Forum on Hybrid AI – GI_Salzburg22

Would you like to go with us on this endeavour and contribute with your expertise?

Then consider joining this Forum, and contact Dr. Lorenz Wendt (lorenz.wendt@plus.ac.at).

The Forum is organised by the Christian Doppler Laboratory GEOHUM, headed by Prof. Stefan Lang (stefan.lang@plus.ac.at).

GI_Salzburg22, formerly AGIT and GI_Forum, provides a platform for dialogue among geospatial minds, informing the GeoInformation Society, contributing to a more just, ethical and sustainable society.

It is the annual meeting place for innovation, networking and continuing education from all areas of Geoinformatics. Interdisciplinary discussions, forward-looking ideas and a personal atmosphere accompany and shape the exchange of science and research, business and application.

GI_Salzburg22 will take place in Salzburg, Austria, 5.-7. July 2022, as a physical, on-site event.

The 12th International Symposium on Digital Earth – GeoHumanitarian Action Forum 8th July 2021

The Department of Geoinformatics – Z_GIS, University of Salzburg hosted the The 12th International Symposium on Digital Earth – ISDE12 – July, 6-8 2021. This special event brought together policymakers and scientists to explore pathways towards the vision of a ‘Digital Earth’.

The GeoHumanitarian Action Forum was a special forum in this year’s ISDE and was held on Thursday, July 8th, 2021. The day’s programme brought together experts from research institutions, service providers, and humanitarian actors, to exchange about the role of geospatial technologies and Earth observation in humanitarian operations. With a mix of presentations, panel discussions, and keynotes, the forum’s interactive format gave both newcomers and experts plenty of opportunity for networking and sharing ideas.

Report – insights and take aways

See the Programme

Call for papers

Geospatial technologies at the interface of satellite Earth observation ( EO ) and geoinformatics ( GI ) are now widely used to support humanitarian operations. The uptake and further development of dedicated information services happens at unprecedented speed, with applications ranging from mission planning and operations in crisis intervention to population estimation for food distribution or vaccination campaigns. The global pandemic experience has even further pushed the need for objective and up-to-date information on impact on mortality rates and mitigation measures.
Co-organized by the Christian-Doppler laboratory GEOHUM, this special forum aims to exchange experience and good practice to comprehend this emerging technology mainly from the following two perspectives:

  • the information needs and technical capacities from humanitarian actors
  • the current trends of cutting-edge technological advancements in the field of EO * GI and artificial intelligence (AI)

We kindly like invite humanitarian organizations (NGOs and GOs), commercial and non-commercial service providers, as well as students and academics active or interested in this field to share their expertise and expectations in a stimulating environment.
Envisaged outcomes of this forum are a better understanding for using and producing relevant information products to optimize logistics, intervention, and mission planning conflict and humanitarian disaster situations.

If you are interested in this forum, you can send your paper proposals to Call for Paper – ISDE12 to share your ideas. Paper submission is possible until the 1st of February 2021.

Opening of the Christian Doppler laboratory GEOHUM (CD laboratory for geospatial and EO-based humanitarian technologies)

Date: Wednesday, 08 July 2020, 15:00 – 17:00 (18:00)
Location: Virtual (via Hopin platform, free registration, main stage)

Content and editorial responsibility
Assoc.-Prof. Dr. Stefan Lang

Programme

15:00 Eröffnungsworte Vizerektorin für Forschung und Nachhaltigkeit Universität Salzburg
Nicola Hüsing
15:05 Grußworte der Bundesministerin für Digitalisierung und Wirtschaftsstandort
Margarete Schramböck
15:10 Grußworte des Präsidenten der Christian Doppler Forschungsgesellschaft
Martin Gerzabek
15:15 Bedeutung des Labors für die Region Salzburg
Landeshauptmann Wilfried Haslauer (Land Salzburg) / Walter Haas (ITG)
15:25 Introduction of the CDL: context, motivation, objectives, overview
Stefan Lang (head of the CDL)
15:45 Motivation of (relevance for) the company partner
Laura Leyser (general director of Ärzte ohne Grenzen, Médecins Sans Frontières Austria – MSF)
Audrey Lessard-Fontaine (head of GIS Unit, Médecins Sans Frontières Suisse – MSF)
16:00 Academic Keynote: “Deep learning for EO applications”
Sébastien Lefèvre, University of South Brittany
16:20 Detailed research topics of the CDL
Lorenz Wendt / Barbara Riedler / Barbara Hofer
16:40 Schlussworte


Master and doctoral theses

PhD Theses
  • Zahra Dabiri: Investigation and evaluation of radar and optical Earth observation data in an object-based image analysis framework [09/2021]
  • Linda Petutschnig: Understanding integrated risks through spatial data [03/2024]
  • Yunya Gao: Applications of deep learning and remote sensing in refugee/internally displaced person (IDP) camp mapping [03/2024]

  • Getachew Gella: Multi-sensor integration for Earth observation based humanitarian action – issus of uncertainty in the information extraction and delivery process [in progress]
MSc Theses
  • Franziska Halbritter: Evaluation of the Regularize Building Footprint function for the generalization of refugee dwellings based on very high resolution satellite imagery [11/2023]
  • Adriko Kennedy: Multi-modal fusion of optical and radar satellite image time series for deforestation mapping in the Amazonian forest [06/2023]
  • Andreas Reithofer: GIS based flood risk assessment of heavy rainfall events for data scarce regions in the case study of Carrefour, Haiti (M.Sc. UNIGIS) [05/2023]
  • Niklas Jaggy: Deep learning powered non-local speckle filtering of Sentinel-1 imagery and its benefits for geohumanitarian action [04/2023]
  • Francesco Collivignarelli: Flood and accessibility monitoring and risk analysis for the humanitarian sector (M.Sc. UNIGIS) [02/2022]
  • Rene Kaspar: Assess the level of damage due to a flood event using a building footprint [11/2021]
  • Tobias Kugler: Höhenmodelle as Multi-Date tristereo Pléiades Satellitenbildern (M.Sc. UNIGIS) [10/2021]
  • Stella Ofori-Ampofo: Multi-sensor crop type identification using pixel-set encoders and temporal self-attention [06/2021]
  • Sheriff Jimoh: Mapping and monitoring slums using geoinformation technologies [06/2021]
  • Frank Willing: Comparison of TempCNN and LightGBM for crop type classification using Sentinel-2 Imagery [01/2021]
  • Emmanouil Skourtis: Coherence-based flood mapping in urban areas (M.Sc. UNIGIS) [ in progress]
  • Johannes Grünewald: Identification of road damages from Sentinel-1 satellite imagery (B.Sc. University of Tübingen) [in progress]
  • Vanessa Streifeneder: Analysing the impact of grey data quality in the training phase of deep-learning models for dwelling extraction in refugee camps

Public outreach

           

Go to Top