Our paper “Spatio-Temporal Diffusion Model for Satellite Imagery” received the Best Paper Award at the Eleventh International Conference on Remote Sensing and Geoinformation of Environment (RSCy 2025).

Starting from sparse satellite observations, our method predicts how landscapes might evolve in the future with changes such as urban expansions or deforested regions, by leveraging a Diffusion Transformer with temporal layers. By training on open-source Landsat and Sentinel-2 data, the model generates future scenarios from a single reference image and a specified target year. This approach provides powerful insights into environmental changes, supporting both climate impact assessments and resource management.

Spatio-Temporal Diffusion Model for Satellite Imagery
Prathap Nagaraj Kashyap, Alireza Javanmardi, Pragati Jaiswal, Alain Pagani, Gerd Reis
RSCy 2025

Congratulations to the authors for this great achievement!

Contact: Alireza Javanmardi, Alain Pagani, Gerd Reis

Dr. Jilliam Díaz Barros, from the Augmented Vision department at DFKI, was honored as an Outstanding Reviewer at WACV 2025.

IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) provides a forum for computer vision researchers working on practical applications and innovative algorithms to share their latest developments.  WACV 2025 was held in Tucson, Arizona, from February 28th to March 4th.

The researchers of the Augmented Vision department have presented 4 papers at the ICPRAM 2025 conference taking place Feb 23 – 25, 2025 in Porto, Portugal.

The International Conference on Pattern Recognition Applications and Methods (ICPRAM) is a point of contact between researchers and engineers working on Pattern Recognition, both from a theoretical and application perspective.

The presented papers are:

“Domain-Incremental Semantic Segmentation for Autonomous Driving under Adverse Driving Conditions”
Shishir Muralidhara, René Schuster, Didier Stricker

“1D-DiffNPHR: 1D Diffusion Neural Parametric Head Reconstruction using a single image”
Pragati Jaiswal, Tewodros Amberbir Habtegebrial, Didier Stricker

“HI^2: Sparse-View 3D Object Reconstruction with a Hybrid Implicit Initialization”
Pragati Jaiswal, Didier Stricker

“Object-Centric 2D Gaussian Splatting: Background Removal and Occlusion-Aware Pruning for Compact Object Models”
Marcel Rogge, Didier Stricker

The paper “HI^2: Sparse-View 3D Object Reconstruction with a Hybrid Implicit Initialization” was a nominee for the best paper award.

Congratulations to all the authors!

Contact: René Schuster, Didier Stricker

The researchers of the department Augmented Vision have presented 4 papers at WACV 2025 conference taking place Feb 28 – Mar 4, 2025 in Tucson, Arizona, USA.

The IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) one of the three major Computer Vision conferences organized by TCPAMI.

The 4 papers are:

“Uni-SLAM: Uncertainty-Aware Neural Implicit SLAM for Real-Time Dense Indoor Scene Reconstruction” Shaoxiang Wang, Yaxu Xie, Chun-Peng Chang, Christen Millerdurai, Alain Pagani, Didier Stricker

“Modality-Incremental Learning with Disjoint Relevance Mapping Networks for Image-based Semantic Segmentation”, Niharika Hegde, Shishir Muralidhara, René Schuster, Didier Stricker

“AnonyNoise: Anonymizing Event Data with Smart Noise to Outsmart Re-Identification and Preserve Privacy”, Katharina Bendig, René Schuster, Nicole Thiemer, Karen Joisten, Didier Stricker

“Beyond Boxes: Mask-Guided Spatio-Temporal Feature Aggregation for Video Object Detection”, Khurram Azeem Hashmi, Talha Uddin Sheikh, Didier Stricker, Muhammad Zeshan Afzal

Congratulations to the authors for this great achievement!

Contact: Alain Pagani, René Schuster, Zeshan Afzal

Contact: Alain Pagani, René Schuster, Zeshan Afzal

The consortium of the HumanTech project, under the coordination of DFKI Augmented Vision – Dr. Jason Rambach,  had its final in-person General Assembly meeting in Madrid, Spain on the 30-31^st January 2025. The meeting was hosted in the premises of the project partner ACCIONA. All project partners had an opportunity to experience live demonstrations of the HumanTech technologies bringing AI and Robots to construction sites, such as intention-activated exoskeletons for construction workers and collaborative brick-laying robots.

The HumanTech project is ending in May 2025 with several success stories such as:

• 5 innovative pilot deployments of HumanTech technologies.
• 15+ scientific publications including top computer vision and robotics journals and conferences.
• 5 awards for DFKI in International challenges in object pose estimation and Scan-to-BIM.
• Building a vibrant community for AI in Construction, the Tech4Construction cluster.

For more on HumanTech: https://humantech-horizon.eu/news/

Contact: Dr. Jason Rambach

Our article “Resolving Symmetry Ambiguity in Correspondence-based Methods for Instance-level Object Pose Estimation” was published in the prestigious Transactions on Image Processing (TIP) Journal. The work is a collaboration of DFKI with Zhejiang University. The article is openly accessible at: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10906413        

Yongliang Lin, Yongzhi Su, Sandeep Inuganti, Yan Di, Naeem Ajilforoushan, Hanqing Yang, Yu Zhang, Jason Rambach. “Resolving Symmetry Ambiguity in Correspondence-based Methods for Instance-level Object Pose Estimation” IEEE Transaction on Image Processing (2025).

Contact: Dr. Jason Rambach

Three EU projects coordinated by the AV department have been presented in a meeting of the Virtual and Augmented Reality Industrial Coalition, hosted by the European Commission in Brussels on February 11^th and 12^th 2025. The meeting brought together over 110 attendees from more than 80 private and public organisations for discussions about the creation of a Private Public Partnership on Virtual Worlds.

The Augmented Vision department presented its work on extended reality (XR) in the projects CORTEX² (Coordinator: Alain Pagani), LUMINOUS (Coordinator:  Zeshan Afzal) and SHARESPACE (Coordinator: Didier Stricker), and participated intensively to the discussions about the Strategic Research and Innovation Agenda (SRIA) of the future Virtual Worlds Partnership.

The topics covered immersion & visualization, intuitive real-time user interaction, standardisation & interoperability, digital twins and XR use cases such as logistics & industry, healthcare, culture media & arts or education.

The department Augmented Vision will continue to represent DFKI in the future partnership on Virtual Worlds with eXtended Reality as key enabling technology.

More information: https://digital-strategy.ec.europa.eu/en/news/virtual-and-augmented-reality-industrial-coalition-remains-committed-solving-real-world-problems

Contact: Alain Pagani, Zeshan Afzal, Didier Stricker

On December 19th, 2024, Jilliam Díaz Barros successfully defended her doctoral thesis entitled ‘Optimization and Generative Models for Face Analysis’. The PhD thesis was carried out in the Augmented Vision department at DFKI, led by Prof. Dr. Didier Stricker, as part of the Computer Science department at RPTU.

The PhD examination commission consisted of Prof. Dr. Didier Stricker (RPTU) and Prof. Dr. Luís Gonzaga Mendes Magalhães (University of Minho) and was chaired by Prof. Dr. Annette Bieniusa (RPTU).

In her thesis, Jilliam Díaz investigated the modelling of the rigid and non-rigid motions of the head, targeting existing challenges in assistance tools and assistive technologies. The PhD thesis focused on four main areas of contributions: head pose estimation, performance-driven facial animation, facial landmark detection and tracking, and face and upper body image synthesis.

Jilliam Díaz received her Bachelor degree in Electronics Engineering from Universidad del Norte, Colombia and her M.Sc. in Computer Science and Electronics from the Université de Bourgogne, France. Besides being part of the Augmented Vision department, Ms. Díaz Barros has been working as software engineer at ANNA Healthcare Saarland UG.

The Network of Excellence dAIEDGE organized two successful workshops at the High Performance Embedded Architectures and Compilers Conference HiPEAC in Barcelona, Spain on January 20-22, 2024.

The workshop “Excellence in AI and Edge Computing” held on Monday, 20 January, was  dedicated to exploring the frontiers of the AI and edge computing synergy, focusing on the development of scalable and robust AI systems. It opened with a keynote by Dr. Alireza Dehghani from the Irish National Centre for Applied AI CeADAR, followed by the presentation of 15 scientific papers and a demo session. The workshop closed with a matchmaking session about “Industrial investment priorities for successful AI at the Edge in Europe”.

The workshop “Charting the Future of Edge AI: Functional and Non-Functional Requirements in the Age of Generative AI” held on Tuesday, 21 January, was organized by Ovidiu Vermesan (SINTEF) and Alain Pagani (DFKI). It provided a comprehensive platform for stakeholders to exchange ideas, share experiences, and collaborate on advancing edge AI technologies, with a special focus on addressing functional and non-functional requirements defined based on system engineering principles. The three sessions covered “Requirements in the Generative AI Era”, “Technology developments” and “Application developments” with presentations, followed by a panel discussion on Edge AI Future Trends.

Coordinator: Dr. Alain Pagani

Project Manager : Dr. Mohamed Selim

Our paper “G3FA: Geometry-guided GAN for Face Animation” was presented at the British Machine Vision Conference (BMVC) 2024 by Alireza Javanmardi from the Augmented Vision Department DFKI!

The British Machine Vision Conference (BMVC) is one of the major international conferences on computer vision and related areas and is organised by the British Machine Vision Association (BMVA). This year it was held in Glasgow from November 25^th to 28^th.

Our work presents a method for real-time face re-enactment from one single source image and addresses a key limitation of GAN-based face animation: geometric consistency.

G3FA enhances realism, especially in challenging head poses, by:

• Utilizing implicit 3D supervision through inverse rendering to extract depth and normal maps from 2D images, guiding the generator towards greater accuracy.
• Employing an ensemble of discriminators to incorporate the 3D properties, improving the generator’s understanding of human head structure.
• Leveraging face volume rendering with orthogonal ray sampling and volume rendering for high-quality image synthesis.

Check out the paper:

G3FA: Geometry-guided GAN for Face Animation, Alireza Javanmardi, Alain Pagani, Didier Stricker, British Machine Vision Conference (BMVC), 2024

Link to the paper

Code is available at github.com/dfki-av/G3FA

Contact: Alireza Javanmardi, Alain Pagani

The Network of Excellence dAIEDGE co-organized the  European Conference on EDGE AI Technologies and Applications (EEAI 2024) in Cagliary, Italy on October 21-23, 2024.

Over the three days, more than 50 presenters took part in 12 keynotes and parallel sessions, addressing more than 20 topics, where participants discussed how Edge AI is redefining the future of technological applications and systems.

dAIEDGE Partners contributed to the EEAI programme with presentations that offered a practical and innovative view of Edge AI in action, highlighting its transformative potential in sectors such as sustainability, energy and security:

FPGA Acceleration for SAR Imagery-Based Sea Ice Mapping: This presentation, by Rashed Al Koutayni (DFKI), addressed the use of FPGA accelerators to improve the speed and accuracy of sea ice mapping using synthetic aperture radar (SAR) imagery.

Functional and Non-Functional Requirements in Edge AI Systems: Ovidiu Vermesan (SINTEF, Norway) and Alain Pagani (DFKI, Germany) discussed the challenges associated with functional and non-functional requirements in Edge AI systems, addressing both operational needs and ethical and legal constraints.

Legal Requirements for Trustworthy Edge AI: Lydia Belkadi from KU Leuven addressed the regulations needed to ensure trustworthy edge AI solutions

The event allowed participants not only to learn from leading experts, but also to network and connect with the best in the industry, strengthening a community of experts and practitioners who are driving innovation in Edge AI.

Coordinator: Dr. Alain Pagani

Project Manager: Dr. Mohamed Selim

The researchers of the Augmented Vision department presented their latest research activities in the field of eXtended Reality at the EuroXR conference in Athens, from November 27^th to 29^th, 2024.

The European Association for eXtended Reality (EuroXR) Conference is the ultimate gathering for Virtual Reality (VR), Augmented Reality (AR), and Mixed Reality (MR). This year, the conference was co-organized by the Institute of Communication and Computer System (ICCS) of the National Technical University of Athens (NTUA).

This year, a joint event on Next Generation XR was co-organized by six European Projects, among them CORTEX2, LUMINOUS and SHARESPACE, all coordinated by the Augmented Vision department at DFKI.

• Alain Pagani presented the recent work on cooperative telepresence in the project CORTEX2
• Michael Lorenz talked about social XR in SHARESPACE
• Zeshan Afzal featured the use of language models for XR in the project LUMINOUS

The coordinators took part in a panel discussion about the “Future of XR”, in which they could share their view about the next steps in XR research and development.

In addition, the AV group presented two application papers and one demonstrator, which attracted the interest of the conference participants

• “Augmented telecommunication in factory setting”, Narek Minaskan, Bastian Krayer, Alain Pagani, Didier Stricker, EuroXR 2024
• “AR Assistant for Pruning of Grapevines and Fruit Trees”, Mariia Podguzova, Simon Haering, Jamiu Ojeleye, Stephan Krauss, Didier Stricker, EuroXR 2024
• “Industrial remote cooperation (Demo)”, Bastian Krayer, Narek Minaskan, Alain Pagani, Didier Stricker, EuroXR

The paper “AR Assistant for Pruning of Grapevines and Fruit Trees” won the EuroXR 2024 Best Application Paper award – Congratulations to the authors!

 
Contact: Dr. Alain Pagani

Dr. Jason Rambach, gave a talk on “Building Virtual Worlds with 3D Sensing and AI” at the NEM Summit 2024 in Brussels on the 23.10.2024. The presentation was part of the AI for Virtual Worlds session organized by Future Artificial Intelligence Research (FAIR). The presentation with results from the Project HumanTech and various XR projects based on object pose estimation technology from DFKI, was followed by an exciting panel discussion.                                                

Contact: Dr. Jason Rambach

The HumanTech project, under the coordination of DFKI Augmented Vision – Dr. Jason Rambach, has reached another key milestone with a highly productive General Assembly Meeting held from 5 to 6 September 2024 at the Technological University of the Shannon (TUS), Athlone Campus, Ireland. As the project is on its final path toward completion next year, this meeting marked an important step in aligning strategies and objectives for the remaining phases.

The overall status of the project shows that everything is progressing as planned, with the pilots well on track and providing a solid foundation for both achieving the current project goals and supporting future research and development initiatives. Additionally, participants enjoyed a guided tour and live demonstration at TUS’s state-of-the-art robotics department, showcasing cutting-edge technology that will play a crucial role in advancing AI applications in the construction industry.

For more on HumanTech: https://humantech-horizon.eu/news/

Contact: Dr. Jason Rambach

We are glad to announce that our paper, Text2CAD: Generating Sequential CAD Models from Beginner-to-Expert Level Text Prompts has been accepted as a spotlight presentation at the Thirty-Eighth Annual Conference on Neural Information Processing Systems (NeurIPS) 2024. The project page is available here. Our research introduces the first AI framework to generate sequential CAD designs from text prompts that include numerical parameters. We look forward to presenting our work in Vancouver, Canada.

Contact: muhammad_zeshan.afzal@dfki.de , mohammad.khan@dfki.de

Dr. Jason Rambach, coordinator of the EU Horizon Project HumanTech , participated in the “Digital Twins for Sustainable Construction” session at the Sustainable Places conference in Luxembourg (23.9-27.9) organized by projects of the Tech4Construction cluster.

Dr. Rambach presented results from the Project HumanTech, focusing on the automated Scan-to-BIM problem. In addition, he was a panel member for the ensuing discussion together with coordinator and members of other ongoing projects in construction and building management such as RobetArme, Beeyonders, InCube and Reincarnate.

Workshop schedule: https://www.sustainableplaces.eu/sp2024/exploring-digital-twins-machine-human-collaboration-and-sustainable-construction/

HumanTech project:     https://humantech-horizon.eu/                                                           

Contact: Dr. Jason Rambach

The Network of Excellence dAIEDGE held its General Assembly meeting on September 5^th and 6th, looking back at the achievements of the last 12 months, and preparing the next steps of the network! It’s been two days full of inspiration and breakthroughs in Heraklion, Crete!

The 36 dAIEDGE partners gathered to share achievements and define the roadmap towards our big goal: To make Europe a world leader in AI and data economy!

During this event, we reinforced our strategic partnerships and strengthened our joint commitment to take technological innovation to another level.

A big thank to our partner ICS-FORTH for being exceptional hosts and making this assembly an unforgettable experience.

We continue to move forward together towards a bright future of innovation and growth!

Coordinator: Dr. Alain Pagani

Project Manager : Dr. Mohamed Selim

We are very happy that our paper “CLEO: Continual Learning of Evolving Ontologies” has been accepted to the European Conference on Computer Vision (ECCV). A preprint of our work can be found here. A short video is available on YouTube. The paper is a joint work with ZF Friedrichshafen AG and describes some results of our collaboration.

We are proud to announce that our paper “RMS-FlowNet++: Efficient and Robust Multi-scale Scene Flow Estimation for Large-Scale Point Clouds” by Ramy Battrawy, René Schuster, and Didier Stricker has been published in the International Journal of Computer Vision (IJCV). The online version of the paper can be found here a preprint is available here. The paper extends our previous work on efficient scene flow estimation in dense point clouds that has been published at ICRA 2022.

The team of the EU Horizon Project HumanTech , consisting of Mahdi Chamseddine and Dr. Jason Rambach from DFKI Augmented Vision as well as Fabian Kaufmann from RPTU Kaiserslautern – department of Civil Engineering, received the 3^rd place prize in the Scan-to-BIM challenge of the (Computer Vision in the Built Environment) CV4_AEC Workshop of the CVPR 2024 conference.

On the 18.6, the team presented their solution and results as part of the workshop program. Scan-to-BIM solutions are of great importance for the construction community as they automate the generation of as-built models of buildings from 3D scans, and can be used for quality monitoring, robotic task planning and XR visualization, among other applications.

HumanTech project:     https://humantech-horizon.eu/

CV4AEC Workshop page: https://cv4aec.github.io/                                                     

Contact: Dr. Jason Rambach  , Mahdi Chamseddine