Team Fraundorfer

3D Vision, Vision for Robotics, Artificial Intelligence (AI), Geometric Deep Learning, Camera Drones

Our research focus is the the area of 3D computer vision including vision for robotics. A specific focus is on the topic of computer vision for camera drones, e.g. for aerial image processing or camera based navigation. We operate a variety of camera drones of different sizes and also operate an indoor flying arena.

NEWS

1 paper accepted to BMVC 2025

Our paper has been accepted to British Machine Vision Conference (BMVC) 2025.

"C³-GS: Learning Context-aware, Cross-dimension, Cross-scale Feature for Generalizable Gaussian Splatting" by Yuxi Hu, Jun Zhang, Kuangyi Chen, Zhe Zhang, Friedrich Fraundorfer

1 paper accepted at ICME 2025

Our paper has been accepted to IEEE International Conference on Multimedia and Expo (ICME) 2025.

"ICG-MVSNet: Learning Intra-view and Cross-view Relationships for Guidance in Multi-View Stereo" by Yuxi Hu, Jun Zhang, Zhe Zhang, Rafael Weilharter, Yuchen Rao, Kuangyi Chen, Runze Yuan, Friedrich Fraundorfer

1 paper accepted at ICRA 2025

Our paper has been accepted to IEEE International Conference on Robotics and Automation (ICRA) 2025.

"EVLoc: Event-based Visual Localization in LiDAR Maps via Event-Depth Registration" by Kuangyi Chen, Jun Zhang, Friedrich Fraundorfer

1 paper accepted at ICPR 2024

Our paper has been accepted to ICPR 2024.

"SAda-Net: A Self-Supervised Adaptive Stereo Estimation CNN For Remote Sensing Image Data" by Dominik Hirner, Friedrich Fraundorfer

1 paper accepted at ECCV 2024

Our paper has been accepted to ECCV 2024.

"GMM-IKRS: Gaussian Mixture Models for Interpretable Keypoint Refinement and Scoring" by Emanuele Santellani, Martin Zach, Christian Sormann, Mattia Rossi, Andreas Kuhn, Friedrich Fraundorfer

1 paper accepted at 3DV 2024

Our paper has been accepted to 3DV 2024.

"HOC-Search: Efficient CAD Model and Pose Retrieval From RGB-D Scans" by Stefan Ainetter, Sinisa Stekovic, Friedrich Fraundorfer, Vincent Lepetit

1 paper accepted at WACV 2024

Our paper has been accepted to WACV 2024.

"HAMMER: Learning Entropy Maps To Create Accurate 3D Models in Multi-View Stereo" by Rafael Weilharter, Friedrich Fraundorfer

1 paper accepted at ICCV 2023

Our paper has been accepted to ICCV 2023.

"S-TREK: Sequential Translation and Rotation Equivariant Keypoints for local feature extraction" by Emanuele Santellani, Christian Sormann, Mattia Rossi, Andreas Kuhn, Friedrich Fraundorfer

2 papers presented at WACV 2023

Our papers were presented at WACV 2023 in Waikoloa.

"Automatically Annotating Indoor Images With CAD Models via RGB-D Scans" by Stefan Ainetter, Sinisa Stekovic, Friedrich Fraundorfer, Vincent Lepetit

"DELS-MVS: Deep Epipolar Line Search for Multi-View Stereo" by Christian Sormann, Emanuele Santellani, Mattia Rossi, Andreas Kuhn, Friedrich Fraundorfer

3 papers accepted at ICPR 2020

October 11, 2020

Our papers have been accepted for ICPR 2020.

"FC-DCNN: A densely connected neural network for stereo estimation" by Dominik Hirner and Friedrich Fraundorfer
"Aerial Road Segmentation in the Presence of Topological Label Noise" by Corentin Henry, Friedrich Fraundorfer, Eleonora Vig
"Machine-learned Regularization and Polygonization of Building Segmentation Masks" by Stefano Zorzi, Ksenia Bittner, Friedrich Fraundorfer

2 papers accepted at 3DV 2020

October 2, 2020

Our papers have been accepted for 3DV 2020.

"BP-MVSNet: Belief-Propagation-Layers for Multi-View-Stereo" by Christian Sormann, Patrick Knöbelreiter, Mattia Rossi, Andreas Kuhn, Thomas Pock, Friedrich Fraundorfer
"DeepC-MVS: Deep Confidence Prediction for Multi-View Stereo Reconstruction" by Andreas Kuhn, Christian Sormann, Mattia Rossi, Oliver Erdler, Friedrich Fraundorfer

Tobias Koch finished PhD thesis

March 26, 2020

My PhD student Tobias Koch finished his PhD thesis with the title "Automated and Precise 3D Building Reconstruction using UAVs" and successfully defended it on March 25th, 2020.

Congratulations.

2 papers accepted at CVPR 2020

February 27, 2020

Our papers have been accepted for CVPR 2020 and will be presented in Seattle.

"Minimal Solutions for Relative Pose with a Single Affine Correspondence" by Banglei Guan, Ji Zhao, Zhang Li, Fang Sun, Friedrich Fraundorfer
"Belief Propagation Reloaded: Learning BP-Layers for Labeling Problems" by Patrick Knöbelreiter, Christian Sormann, Alexander Shekhovtsov, Friedrich Fraundorfer, Thomas Pock

ICRA 2019 paper accepted

January 29, 2019

Our paper "RESLAM: A real-time robust edge-based SLAM system" has been accepted and will be presented at ICRA 2019 in Montreal, Canada.

JFR paper accepted

January 24, 2019

Our paper "Overview Obstacle Maps for Obstacle Aware Navigation of Autonomous Drones" has been accepted for publication in the Journal of Field Robotics (JFR).

Post delivery drone featured at the Vienna Museum of Technology

November 16, 2018

Our post delivery drone prototype is being featured at the exhibition "Arbeit & Produktion" at the Vienna Museum of Technology (Technisches Museum Wien). Read more information about it in this news article, and checkout our picture of the drone [JPG].

Search and rescue drones project featured in documentary film

October 18, 2018

Our work on search and rescue drones got featured in the TV program P.M. Wissen. The documentary film about the topic of search and rescue drones, of which our work was a part of, was aired the 18.10.18 by ServusTV in the TV program P.M. Wissen.

Poster presentation at IROS2018 Logistics Workshop

October 5, 2018

We presented a poster about our Inventory Drone feasibility project prototype in the IROS 2018 Workshop - Robotics for logistics in warehouses and environments shared with humans - in Madrid, Spain the 5th October, 2018. Our extended abstract entitled "Towards an Autonomous Vision-based Inventory Drone" is available: [PDF], [REFILLS Project - Events] and [PDF]. The full program of the workshop is available [IROS2018W Programme] and [REFILLS Project - Events].

ECCV paper accepted

July 11, 2018

Our paper "Semantically Aware Urban 3D Reconstruction with Plane-Based Regularization" [PDF] [Supp. Material] [video] has been accepted and will be presented at ECCV 2018 in Munich!

PaperPresented at UAVg 2017

September 08, 2017

We presented our paper "AUTOMATED INSPECTION OF POWER LINE CORRIDORS TO MEASURE VEGETATION UNDERCUT USING UAV-BASED IMAGES" [pdf] at the UAVg 2017 at Bonn from September 04th-07th.

BMVC 2017 Papers Accepted

July 24, 2017

Our papers "Combining Edge Images and Depth Maps for Robust Visual Odometry" [pdf] and "Plane-based Surface Regularization for Urban 3D Reconstruction" have been accepted and will be presented at this years' BMVC from 4th -7th September at Imperial College London.

Delivery drone project in the news

July 10, 2017

Our work on delivery drones got features in the news. Nice project video is available (Video)

Book Chapter: Computer Vision for MAVs!

March 20, 2017

The book "Computer Vision in Vehicle Technology: Land, Sea, and Air" is now available online and as printed book. It contains a chapter about "Computer Vision for MAVs" written by Friedrich Fraundorfer.

Awarded with Amazon gift !

March 1, 2017

Our team got awarded an Amazon gift for continuing our drone related research.

CVPR 2017 paper accepted!

February 28, 2017

Our paper has been accepted for publication at CVPR 2017! Scalable Surface Reconstruction from Point Clouds with Extreme Scale and Density Diversity. Christian Mostegel, Rudolf Prettenthaler, Friedrich Fraundorfer, Horst Bischof.

3DV'16 paper accepted!

September 15, 2016

Our paper "Regularized 3D Modeling from Noisy Building Reconstructions" has been accepted for publication at 3DV'16! You can download a preprint version from [here].

Presentation at Ars Electronica Festival 2016

September 8-12, 2016

We presented our "Intelligent Search and Rescue Drone" in the Ars Electronica Festival 2016 - DroneLab, Linz, September 8-12, 2016. This was a great opportunity to approach the general public about civilian applications for drones and research made in the ICG (TU Graz). We used this opportunity to show our drone and the capabilities of our 3D resconstruction software, and to describe the Graz Griffins', our team, participation in the 2016 DJI Developer Challenge.

We qualified for DJI Developer Challenge Round 3!

August 9, 2016

Our team, the Graz Griffins is one of 10 teams which qualified for the 3rd round of the DJI Developer Challenge2016 (Team Listings)! Hence, we are allowed to participate in the final competition end of August at the Griffiss International Airport, Rome, New York, USA. [more]

CVPR Code Released!

May 19, 2016

We released our code of our CVPR paper, with which you can automatically generate training data for confidence learning in stereo vision. You can find it in the "Software" section.

CVPR and CVPRW paper accepted!

May 09, 2016

We are happy to announce that two of our papers got accepted to CVPR this year. One paper (Using Self-Contradiction to Learn Confidence Measures in Stereo Vision) will be presented at the main conference and a second paper (UAV-based Autonomous Image Acquisition with Multi-View Stereo Quality Assurance by Confidence Prediction) at the CVPR workshop on Computer Vision in Vehicle Technology.

Journal paper accepted!

Apr 05, 2016

We are happy to announce that our journal paper got accepted at Computer Vision and Image Understanding (CVIU). More details can be found in the Publications section.

DJI Developer Challenge 2016!

Mar 29, 2016

Our team, the Graz
Griffins, has qualified to participate in the 2016 DJI Developer Challenge (see Challengers >> 1st round), among only 25 teams worldwide. The team consists of members from the IRT and the ICG, both institutes at the TU Graz. We will now receive our Developer Package, which includes the DJI Matrice M100 quadcopter equipped with the impressive DJI Guidance multipurpose sensing system, a Gimbal Zenmuse X3 camera and a DJI Manifold onboard computer!

VISAPP'16: Best Paper Award!

Mar 03, 2016

We won the Best Paper Award for our paper "Visual Odometry Based on Lines", which has been presented at this years VISAPP.

Direct Stereo Visual Odometry Based on Lines

Jan 22, 2016

Our new paper Direct Stereo Visual Odometry Based on Lines is now online, and can be found in the Publications section. It will be presented at this years VISAPP in Rome.