Torsten Sattler
RICAIP Tenure Track holder
Senior researcher at the Czech Institute of Informatics, Robotics and Cybernetics (CIIRC)
Computer Vision – 3D Reconstruction – Visual Localization
Previous Experience
With his work, Dr Torsten Sattler has contributed to the Institute of Visual Computing at one of the world’s leading universities in the field of technology and natural sciences, ETH Zurich, Switzerland. He graduated in computer science and got his doctoral degree at RWTH Aachen, the largest university of technology in Germany and one of the most renowned in Europe. Before joining CIIRC, Dr Torsten Sattler was an associate professor at Chalmers Technical University in Gothenburg, Sweden.
His main research interest is 3D computer vision, with a special focus on 3D reconstruction and visual localization. At CIIRC CTU, he is currently leading a team of 8 PhD students working on machine learning and 3D Computer Vision.
I wish RICAIP all the best and all success for its next stage.
Current Focus
Torsten’s work explores interconnections of computer vision, robotics, augmented reality, and AI. He focuses on advanced localization and 3D reconstruction algorithms that work robustly and reliably on a wider range of scenes while modelling the fact that the real world is dynamic.
Torsten Sattler contributes to the Use-Cases development with the knowledge about 3D computer vision and machine learning for 3D vision to help automate industrial tasks. His work on mapping and localization has applications in both robotics and augmented reality systems, such as digital twins. His current research with a PhD student at the RICAIP Testbed Prague focuses on leveraging 3D reconstruction for tasks such as object pose estimation, enabling more robust and accurate object manipulation by robots using RGB cameras. He is the Principal Investigator and main beneficiary in the project A Unified 3D Map Representation (GACR EXPRO, 2023-2027).
Recent successes and results – What recent research achievements or milestones are you most proud of?
I am probably most proud of being part of a work (Zehao Yu, Anpei Chen, Binbin Huang, Torsten Sattler, Andreas Geiger, Mip-Splatting: Alias-free 3D Gaussian Splatting) that received the Best Student Paper Award at CVPR 2024. CVPR is one of the main conferences in the field of computer vision, with typically more than 10,000 submissions. Being part of an award winning work is a great honor.
Biggest challenges and future plans – What are the main challenges currently facing your research areas, and what directions do you plan to pursue in the coming years?
Given the hunger for more and more data in machine learning, I think privacy-preserving visual perception will become more and more important. I.e., rather than having to send full images to external servers, it is preferable to send only abstract images, e.g., segmentations, that reveal as little as possible about the scene and thus about a user’s environment. As more and more robots are applied in households, being able to preserve user privacy will become more and more important.
International collaborations with other teams/ projects etc – Your work involves collaborations across institutions and projects. How do international partnerships influence your research, and what value do they bring?
International collaborations are a great way to expand the range of topics I am working on and to get fresh perspectives on research problems. An example for the latter is a collaboration with the University of Tübingen, which led to the paper award mentioned above. I currently have ongoing collaborations with ETH Zurich, Comenius University, the University of Amsterdam, and Naver Labs Europe. Multiple of these collaborations are through the ELLIS PhD program, where some of my students have co-advisors in other European countries. In general, I believe that international partnerships are invaluable, especially when it comes to keeping Europe internationally competitive as a way to harness all the talent distributed over Europe.
Selected publications
| Charalambos Tzamos, Viktor Kocur, Yaqing Ding, Torsten Sattler, Zuzana Kukelova Are Minimal Radial Distortion Solvers Necessary for Relative Pose Estimation? arxiv | |
| Charalambos Tzamos, Viktor Kocur, Yaqing Ding, Daniel Barath, Zuzana Berger Haladová, Torsten Sattler, Zuzana Kukelov Practical solutions to the relative pose of three calibrated camera Proceedings of the Computer Vision and Pattern Recognition Conference | |
| Kunal Chelani, Assia Benbihi, Fredrik Kahl, Torsten Sattler, Zuzana Kukelova Obfuscation Based Privacy Preserving Representations are Recoverable Using Neighborhood Information arxiv | |
| Vaclav Vavra, Torsten Sattler, Zuzana Kukelova Camera Pose Estimation from Bounding Boxes 2024 IEEE/RSJ | |
| Varun Burde, Assia Benbihi, Pavel Burget and Torsten Sattler Comparative Evaluation of 3D Reconstruction Methods for Object Pose Estimation arXiv preprint, 2024 | |
| Tzamos, Charalambos; Barath, Daniel; Sattler, Torsten; Kukelova, Zuzana Relative pose of three calibrated and partially calibrated cameras from four points using virtual correspondences arXiv preprint, 2024 | |
| Jonathan Ventura, Zuzana Kukelova, Torsten Sattler, Dániel Baráth Absolute Pose from One or Two Scaled and Oriented Features Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2024 | |
| Jonathan Ventura, Zuzana Kukelova, Torsten Sattler, Daniel Barath P1AC: Revisiting Absolute Pose From a Single Affine Correspondence. International Conference on Computer Vision (ICCV) 2023. | |
| Kunal Chelani, Torsten Sattler, Fredrik Kahl, Zuzana Kukelova Privacy-Preserving Representations Are Not Enough: Recovering Scene Content From Camera Poses, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2023 | |
| S Bhayani, V Larsson, T Sattler, J Heikkila, Z Kukelova Partially calibrated semi-generalized pose from hybrid point correspondences. IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2023. | |
| G Berton, R Mereu, G Trivigno, C Masone, G Csurka, T Sattler, B Caputo Deep Visual Geo-Localization Benchmark. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) | |
| M Humenberger, Y Cabon, N Pion, P Weinzaepfel, D Lee, N Guérin, T Sattler, G Csurka Investigating the Role of Image Retrieval for Visual Localization. International Journal of Computer Vision. | |
| A Adam, T Sattler, K Karantzalos, T Pajdla Objects Can Move: 3D Change Detection by Geometric Transformation Consistency. European Conference on Computer Vision (ECCV), 2022. | |
| V Panek, Z Kuikelova, T Sattler MeshLoc: Mesh-Based Visual Localization. European Conference on Computer Vision (ECCV), 2022. | |
| J Kulhánek, E Derner, T Sattler, R Babuska ViewFormer: NeRF-free Neural Rendering from Few Images Using Transformers. European Conference on Computer Vision (ECCV), 2022. | |
| Z Yu, S Peng, M Niemeyer, T Sattler, A Geiger MonoSDF: Exploring Monocular Geometric Cues for Neural Implicit Surface Reconstruction. 2022 Conference on Neural Information Processing Systems NeurIPS | |
| A Jafarzadeh, M López Antequera, P Gargallo, Y Kuang, C Toft, F Kahl, T Sattler: CrowdDriven: A New Challenging Dataset for Outdoor Visual Localization IEEE International Conference on Computer Vision (ICCV) (2021) | |
| S Bhayani, T Sattler, D Barath, P Beliansky, J Heikkila, Z Kukelova: Calibrated and Partially Calibrated Semi-Generalized Homographies IEEE International Conference on Computer Vision (ICCV) (2021) | |
| E Brachmann, M Humenberger, C Rother, T Sattler: On the Limits of Pseudo Ground Truth in Visual Camera Re-localisation IEEE International Conference on Computer Vision (ICCV) (2021) | |
| V Guzov, A Mir, T Sattler, G Pons-Moll: Human POSEitioning System (HPS): 3D Human Pose Estimation and Self-localization in Large Scenes from Body-Mounted Sensors IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (2021) (oral, best paper candidate) | |
| P-E Sarlin, A Unagar, M Larsson, H Germain, C Toft, V Larsson, M Pollefeys, V Lepetit, L Hammarstrand, F Kahl, T Sattler: Back to the Feature: Learning Robust Camera Localization from Pixels to Pose IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (2021) | |
| K Chelani, F Kahl, T Sattler: How Privacy-Preserving are Line Clouds? Recovering Scene Details from 3D Lines IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (2021) (oral) | |
| Q Zhou, T Sattler, L Leal-Taixe: Patch2Pix: Epipolar-Guided Pixel-Level Correspondences IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (2021) | |
| A Sunegard, L Svensson, T Sattler: Deep LiDAR localization using optical flow sensor-map correspondences IEEE International Conference on 3D Vision (3DV) (2020) | |
| E Stenborg, T Sattler, L Hammarstrand: Using Image Sequences for Long-Term Visual Localization IEEE International Conference on 3D Vision (3DV) (2020) | |
| N Pion, M Humenberger, G Csurka, Y Cabon, T Sattler: Benchmarking Image Retrieval for Visual Localization IEEE International Conference on 3D Vision (3DV) (2020) | |
| Z Zhang, T Sattler, D Scaramuzza: Reference Pose Generation for Long-term Visual Localization via Learned Features and View Synthesis, International Journal of Computer Vision, Special Issue on Performance Evaluation in Computer Vision, 2020. |
