Our professorship is thrilled to announce its active participation in the 31st International Conference on Advances in Geographic Information Systems (SIGSPATIAL ‘23), taking place in Hamburg, Germany. We are co-organizing the conference, are co-chairing multiple workshops, and we are presenting quite a few research results from colleagues of the chair and our collaborators.

Research Papers

We are excited to present the following scientific papers:

1. Rethink Geographical Generalizability with Unsupervised Self-Attention Model Ensemble: A Case Study of OpenStreetMap Missing Building Detection in Africa (full paper) Authors: Hao Li, Jiapan Wang, Johann Maximilian Zollner, Gengchen Mai, Ni Lao, Martin Werner

2. Bavaria Buildings - A Novel Dataset for Building Footprint Extraction, Instance Segmentation, and Data Quality Estimation (Data and Resources Paper) Authors: Martin Werner, Hao Li, Johann Maximilian Zollner, Balthasar Teuscher, Fabian Deuser

3. Exploring GeoAI Methods for Supraglacial Lake Mapping on Greenland Ice Sheet (GIS cup) Authors: Xuanshu Luo, Paul Walther, Wejdene Mansour, Balthasar Teuscher, Johann Maximilian Zollner, Hao Li, Martin Werner

4. Signal Separation in Global, Temporal Gravity Data (GeoAI workshop paper) Authors: Betty Heller-Kaikov, Roland Pail and Martin Werner

5. Towards GeoAI as a Containerized Microservice (SRC paper) Authors: Jiapan Wang

Workshops

We are delighted to co-host two workshops:

1. 11th ACM SIGSPATIAL International Workshop on Analytics for Big Geospatial Data (BigSpatial 2023) Workshop Co-Chairs from our group: Martin Werner Description: Join us to explore the challenges and opportunities of processing and analyzing big geospatial data, fostering collaboration and knowledge exchange.

2. 2nd ACM SIGSPATIAL International Workshop on Searching and Mining Large Collections of Geospatial Data Workshop Co-Chairs from our group: Hao Li, Martin Werner Description: Engage with experts in the efficient searching and mining of large geospatial data collections, contributing to the development of cutting-edge solutions.

For further information about our participation and research reach out to us on the conference. We look forward to sharing insights, learning, and contributing to the vibrant academic community at SIGSPATIAL ‘23. For detailed information on all our published papers refer to the publications page.

Our PhD student Michael Petry has prepared a four minute thesis presentation on his Phd project Onboard Machine Learning for Physical Layer Signal Processing on Telecom-Satellites for the Four-Minute-Thesis (4MT) Competition hosted by IEEE Globecom.

In this challenge, PhD students are invited to submit a max. 4 minute long video explaining their PhD topic to a non-technical audience. In this video, Michael is explaining how he tries to implement software-defined radio systems using mainly (or only?) neural networks thus enabling AI specific hardware systems such as Xilinx Versal FPGAs to take over more aspects of the communication system.

In cooperation with the Chair of Soil Science at the Technical University of Munich, the Professorship Big Geospatial Data Management hosted a hackathon as part of the project SoilCarbonHack.

We gathered soil-driven and data-driven minds to work together on microscale NanoSIMS images and improve our understanding of soil carbon storage.

We had new ideas, engaging discussions, and could benefit from each others field knowledge during this interdisciplinary meeting.

If you are interested in the hands-on tutorials, we provide the related Jupyter Notebooks and an example NanoSIMS image on our hackathon page.

Stay tuned for upcoming events and papers from SoilCarbonHack!

For information on the project and upcoming events please refer to the project page.

Hao Li was presenting results of our efficient point cloud query paper (Teuscher et al., 2023) led by Balthasar Teuscher at the 18th 3DGeoInfo 2023 conference in Munich.

In this paper, we propose an efficient in-memory point cloud processing solution and implementation demonstrating that the inherent technical identity of the memory location of a point (e.g., a memory pointer) is both sufficient and elegant to avoid gridding as long as the point cloud fits into the main memory of the computing system. During the conference, we have collected a handful of nice comments and suggestions for the participants, which will be integrated in the future development. This paper is a nice joint effort with TUM colleagues from Chair of Engineering Geodesy (Prof. Holst) and Professorship for Remote Sensing Applications (Prof. Anders).

Resources

• Github Repository

1. Teuscher, B., Geißendörfer, O., Luo, X., Li, H., Anders, K., Holst, C., & Werner, M. (2023). Efficient In-Memory Point Cloud Query Processing. 18th International 3DGeoInfo Conference 2023. [PDF] [BibTeX]

   @article{2023_pointcloudqueries_Teuscher,
     author = {Teuscher, Balthasar and Geißendörfer, Oliver and Luo, Xuanshu and Li, Hao and Anders, Katharina and Holst, Christoph and Werner, Martin},
     title = {Efficient In-Memory Point Cloud Query Processing},
     journal = {18th International 3DGeoInfo Conference 2023},
     year = {2023}
   }
   

Carla Rieger was a speaker in this year’s CERN Openlab summer student programme, where she presented on Quantum Machine Learning and Quantum Optimization. Furthermore, her talk was showcasing successful CERN use-cases to illustrate current practical applications of quantum algorithms in the field.

More details, including the lecture slides and a recording of the talk, can be found here.

The professorship “Big Geospatial Data Management” is seeking to fill a research assistant (E13) position to support research activities around geospatial artificial intelligence and spatial data management with a view toward modern didactics, starting as soon as possible.

TUM Portal Link

PDF Version

Register

The registration for the hackathon as part of the project SoilCarbonHack is now open!

To participate, please fill in the form here.

Hackathon

Detailed information on the project and event can be found here.

We are proud to announce that a team led by Fabian Deuser won the first place at the cross-view geolocalization competition.

In this challenge, the organizers presented a novel, challenging cross-view geo-localization dataset, called University160k. The motivation was to provide a comparably large satellite-view dataset for geolocalization to increase the number of similar features in different images. With a strategy of using pseudolabels to get a good alignment of latent space features and the localization problem, we were able to outperform all other submissions in the challenge. A paper on this topic is accepted and will be presented in the workshop.

Congratulations to our fresh (first month?) PhD student Fabian Deuser, who was the lead in all of this work.

This challenge has triggered a new line of research in our group as we believe that the true geolocalization problem is even harder than already depicted in the enlarged University160k dataset. On the other hand, the localization problem is typically a local problem as a coarse location might already be known in most applications. Furthermore, we will extend this activity to the indoor space, where even a limited-scalability reliable indoor geolocalization from images would be very helpful.

So stay tuned for the workshop presentation, the paper, and our follow-up work maybe including additional geolocalization challenges.

In cooperation with the Chair of Soil Science at the Technical University of Munich, the Professorship Big Geospatial Data Management is hosting a hackathon as part of the project SoilCarbonHack.
Please refer to the project page for detailed information on the project and events.

When?

12.10.2023 - 13.10.2023

Where?

Room 0120 in 0501 Institutsbau
Technical University of Munich
Arcisstraße 21
80333 München

Contact

Johann Maximilian Zollner
maximilian.zollner@tum.de
Professorship of Big Geospatial Data Management
Lise-Meitner-Str. 9
85521 Ottobrunn

Yahan Hu
yahan.hu@tum.de
Chair of Soil Science
Emil-Ramann-Straße 2
85354 Freising

On July 13, the room was packed with colleagues and students when Wenwen Li revealed to us her experience from adapting vision foundation models to geospatial images. It was very interesting for all of us and as the discussion grew into a long session shows how important this topic is.

In order to sustain the exchange and to turn the frontal oneway communication of such a presentation format into a more valuable interactive exchange, we organize a follow-up and lessons learnt session in three months.

Ressources

• Archiv Preprint on Permafrost Feature Example
• Google Scholar Page

A novel teaching aid in our lab. The Xilinx AI System KV260 provides all you need for embedded and mobile AI applications. Full FPGA board with a small form factor. Easy to use as it runs a Ubuntu Linux.

Students will learn more about mobile AI with this excellent board.

Hao Li was presenting aspects of our AtlasHDF infrastructure (Werner & Li, 2022) which was explored for surface water segmentation in the context of NFDI4Earth.

In a nutshell, we propose a shift in geospatial big data from Geo to NoGeo quite in the same way as the big data community sacrificed the SQL language in big data towards NoSQL infrastructures. OGC standards and their libraries are valuable assets for data management, interpretation, and preservation. But they have never been designed for computation. While our approach is less powerful in spatial operations at the moment, we have zero dependencies beyond HDF5. But this is mature, stable, and - most importantly - an integral part of both tensorflow and pytorch. The philosophy of this approach is simplicity: our data can be used by every deep learning scientist out of the box. And the best: supercomputers read it in parallel…

Ressources

• Github Repository
• Github Mirror at End of Project

1. Werner, M., & Li, H. (2022). AtlasHDF: An Efficient Big Data Framework for GeoAI. 1–7. https://doi.org/10.1145/3557917.3567615 [PDF] [Online] [BibTeX]

   @inproceedings{2022_atlashdf_Werner,
     author = {Werner, Martin and Li, Hao},
     title = {AtlasHDF: An Efficient Big Data Framework for GeoAI},
     year = {2022},
     isbn = {9781450395311},
     publisher = {Association for Computing Machinery},
     address = {New York, NY, USA},
     url = {https://doi.org/10.1145/3557917.3567615},
     doi = {10.1145/3557917.3567615},
     pages = {1–7},
     numpages = {7},
     keywords = {GeoAI, OpenStreetMap, big data, hierarchical data format, immediate mapping},
     location = {Seattle, Washington},
     series = {BigSpatial '22}
   }
   

The professorship “Big Geospatial Data Management” is looking for academic staff to support developments around artificial intelligence for spatial data management with a view towards hardware design, edge computing, autonomous flight of drones and modern didactics, starting as soon as possible.

TUM Portal Link

PDF Version

The new semester sees the arrival of the very first students at the TU Munich to study our newly established Bachelor Aerospace. We participate in this study program with the two computer science lectures (“Computational Foundations I”, “Computational Foundations II”) and teach the basics of algorithms, programming, and technical programming, and computer engineering, so that our students can develop mission critical software. In times of digitalization, there is probably no subject in which the long-term theoretical foundation (computer science as the foundation of all data science) is as important as actual practical skills (software as a universal tool). Accordingly, we adopt the didactic approach 4CID, which emphasizes learning through practical tasks in in a procedural way (step by step: increasing complexity with decreasing assistance) and thematic embedding (application examples). (examples of application) is in the spotlight.

We are looking forward to this new lecture series, which will be in parts also be used in the next semester in the course of geodesy and geoinformatics.

This year, we again participated in the MINTEntdeckerinnen program of the TU Munich. Six young women between the ages of 15 and 18 were our guests during the school vacations to gain experience in our field.

This year, we selected “autonomous aerial drones” as the topic and learned and tried out all the necessary elements needed to simulate autonomous drones and develop the necessary software.

We learned about MATLAB and SimuLink, got to grips with the Unity game engine, implemented a simulation of drone flight dynamics in MATLAB and connected it to the game engine. In the end we were able towe were able to plan trajectories, fly off, and solve a safe-and-rescue scenario.

One or two drones crashed in the simulation or flew through houses, but in the end all six students solved the problem independently and practically.

At the annual CODE conference at Universität der Bundeswehr München, our submission “CyberSuche” won 2nd prize.

We are happy to announce the following guest lecture:

Friday, July 31st,2020 13:00 – 15:00 CET.

In order to attend, you need to register for this lecture.

Registration Information

• https://tum-conf.zoom.us/meeting/register/tJUvdeCvpjkuE9EF1htysTrW095SR8Jdtxe1

Speaker

Martin Szugat

Abstract

Companies are drowning in data, but are thirsty for information. Although e.g. ESA’s Copernicus Open Data Strategy has exploded the availability & quality of geospatial data, only a small number of companies are using it. Because most companies only see their own data instead of seeing the opportunities and looking for new data. A data strategy is the business plan for data & analytics and data thinking is the method to develop this business plan. In this lecture you will get to know some useful tools to design data products yourself.

Vita

Martin Szugat is the founder and managing director of Datentreiber, a data strategy consulting firm. For his projects e.g. for Roche, ProSiebenSat1 and many more, he applies design thinking to data science and has developed a method and open source tools for data strategy design. Besides he is the program director of the Predictive Analytics & Deep Learning World conferences in Europe. He studied bioinformatics at LMU & TUM. When he has free time, he devotes himself to AI and Space Data and tweets about it on http://acceleran.do. The presentation is mandatory for students of the lecture Big Geospatial Data and open for all interested guests.

Today, I am starting my professorship for big geospatial data management. In these days, a virus is affecting all of us and in this situation it is unfortunately not possible to do a large party. But we will have this party as soon as we can meet in my Big Geoespatial Data Management lab.

For now, we concentrate on excellent teaching and research in times of home-office and online courses. .

Valentines day is a very dominant hashtag. The following shows a very short excerpt of the Twitter 1% public stream. Look at how Valentines day dominates hashtag distribution for this day.