Profile cover photo
Profile photo
Torsten Sattler
158 followers
158 followers
About
Posts

Post has attachment
In case you are interested, the slides from Eric Brachmann's and mine tutorial on Visual Localization, given at ECCV 2018, are online: https://sites.google.com/view/visual-localization-eccv-2018/home
Add a comment...

Post has attachment
Interested in the combination of geometric and semantic scene understanding, e.g., for 3D reconstruction and object detection? Then you might be interested in attending the ECCV 2018 workshop "3D Reconstruction meets Semantics".

We are going to have three invited talks by three great researchers, Andrew Davison, Thomas Funkhouser, and Christian Häne. In addition, there will be spotlight and poster presentations. The workshop will conclude by a panel discussion with the invited speakers.

The workshop takes place in the morning of September 9th.

Please find the detailed schedule here:
http://trimbot2020.webhosting.rug.nl/events/3drms/date-schedule/

For general information, please see
http://trimbot2020.webhosting.rug.nl/events/3drms/
Add a comment...

Post has attachment
If you are interested in visual localization, e.g., for Augmented / Virtual / Mixed Reality, self-driving cars, or robotics, Eric Brachmann and I will be giving a tutorial on this topic at ECCV this year.

The tutorial will consist of three parts: The first part covers classical feature-based methods. The second part covers recent learning-based methods, both for direct pose regression and for scene coordinate regression. The last part discusses failure cases and open problems of both (feature-based and learning-based) approaches. Besides providing details on state-of-the-art localization systems, the goal of the tutorial is to show when different types of systems are preferable.

What: Visual Localization - Feature-based vs. Learned Approaches
When: Saturday, September 8th, 2018
Where: N1070ZG, Technische Universität München Arcisstraße 21, 80333 Munich
Time: 9:00 - 12:40

More details: https://sites.google.com/view/visual-localization-eccv-2018/home
Add a comment...

Post has attachment
Interested in building databases of 3D objects, e.g., for robotics? Scanning each object individually takes too much time? Then you might be interested in our work on automatically building a database of 3D objects while traversing a scene with an RGB-D camera.

The following paper, to be presented at IROS 2018, describes our system:

Fadri Furrer, Tonci Novkovic, Marius Fehr, Abel Gawel, Margarita Grinvald, Torsten Sattler, Roland Siegwart, and Juan Nieto
Incremental Object Database: Building 3D Models from Multiple Partial Observations

Abstract:
Collecting 3D object datasets involves a large amount of manual work and is time consuming. Getting complete models of objects either requires a 3D scanner that covers all the surfaces of an object or one needs to rotate it to completely observe it. We present a system that incrementally builds a database of objects as a mobile agent traverses a scene. Our approach requires no prior knowledge of the shapes present in the scene. Object-like segments are extracted from a global segmentation map, which is built online using the input of segmented RGB-D images. These segments are stored in a database, matched among each other, and merged with other previously observed instances. This allows us to create and improve object models on the fly and to use these merged models to reconstruct also unobserved parts of the scene. The database contains each (potentially merged) object model only once, together with a set of poses where it was observed. We evaluate our pipeline with one public dataset, and on a newly created Google Tango dataset containing four indoor scenes with some of the objects appearing multiple times, both within and across scenes.

Video:
https://www.youtube.com/watch?v=9_xg92qqw70

ArXiv page:
https://arxiv.org/abs/1808.00760
Add a comment...

Post has attachment
Interested in visual localization, e.g., for figuring out where your robot is or for Augmented / Mixed / Virtual Reality? You would like to localize against larger maps, but you are running out of memory? Then you might be interested in the following paper. In it, we propose a new map compression scheme that stores full local descriptors for a tiny subset of all points and (heavily) quantized descriptors for a larger subset of points. We then show how the quantized descriptors can be used to improve pose accuracy and localization rates.

+Fede Camposeco Paulsen +Andy Cohen +Marc Pollefeys +Torsten Sattler
Hybrid scene Compression for Visual Localization

abstract:
Localizing an image wrt. a large scale 3D scene represents a core task for many computer vision applications. The increasing size of available 3D scenes makes visual localization prohibitively slow for real-time applications due to the large amount of data that the system needs to analyze and store. Therefore, compression becomes a necessary step in order to manage large scenes. In this work, we introduce a new hybrid compression algorithm that selects two subsets of points from the original 3D model: a small set of points with full appearance information, and an additional, larger set of points with compressed information. Our algorithm takes into account both spatial coverage as well as appearance uniqueness during compression. Quantization techniques are exploited during compression time, reducing run-time wrt. previous compression methods. A RANSAC variant tailored to our specific compression output is also introduced. Experiments on six large-scale datasets show that our method performs better than previous compression techniques in terms of memory, run-time and accuracy. Furthermore, the localization rates and pose accuracy obtained are comparable to state-of-the-art feature-based methods, while using a small fraction of the memory.

Paper:
https://arxiv.org/abs/1807.07512
Add a comment...

Post has attachment
For those of you interested in semantic scene understanding and 3D modeling, we have extended the paper and challenge submission deadlines for the "3D Reconstruction meets Semantics" workshop held at ECCV 2018 (http://trimbot2020.webhosting.rug.nl/events/3drms/).

Both deadlines are now July 17th (23:59 GMT).
Add a comment...

Post has attachment
Apologies in case you receive this announcement multiple times.

**********************************

CFP: 3DRMS - 3D Reconstruction meets Semantics - ECCV 2018 Workshop

- A Workshop, a Challenge, and peer-reviewed Papers

The Second Workshop on 3D Reconstruction Meets Semantics:
Integration of 3D Vision with Recognition and Learning

**********************************
Date: 9th September 2018 (Morning/Half-day)
Location: Munich, Germany
Website:http://trimbot2020.webhosting.rug.nl/events/3drms/
**********************************


**********************************
IMPORTANT DATES:
**********************************
Submission of full papers and extended abstracts: JULY 10TH, 2018 - 23:59 (GMT)
Submission of challenge results: JULY 10TH, 2018 - 23:59 (GMT)
Notification of acceptance: JULY 31TH, 2018
Challenge results evaluated: AUGUST 1st, 2018
Camera-ready manuscripts: LATE SEPTEMBER, 2018 (after workshop)

**********************************
INVITED SPEAKERS
**********************************
Andrew Davison, Imperial College London, UK
Thomas Funkhouser, Princeton University, USA
Christian Häne, University of California at Berkeley, USA


**********************************
SCOPE:
**********************************
The goal of this workshop is to explore and discuss new ways for
integrating techniques from 3D reconstruction with recognition and
learning. How can semantic information be used to improve the dense
matching process in 3D reconstruction techniques? How valuable is 3D
shape information for the extraction of semantic information? In the age
of deep learning, can we formulate parts of 3D reconstruction as a
learning problem and benefit from combined networks that estimate both
3D structures and their semantic labels? How do we obtain feedback-loops
between semantic segmentation and 3D techniques that improve both
components? Will this help recover more detailed 3D structures?

Topics of interest for this workshop include, but are not limited to:
* Semantic 3D reconstruction and semantic SLAM
* Learning for 3D vision
* Fusion of geometric and semantic maps
* Label transfer via 3D models
* Datasets for semantic reconstruction including synthetic dataset generation for learning
* 2D/3D scene understanding and object detection
* Joint object segmentation and depth layering
* Correspondence and label generation from semantic 3D models
* Robotics applications based on semantic reconstructions
* Semantically annotated models for augmented reality

We encourage both the submission of original work in the form of full
papers and work in progress in the form of an extended abstract. Full
papers will be peer-reviewed and will be published in the proceedings of
the workshop through Springer LNCS. Extended abstracts will not be
formally published, but we will collect the abstracts on the website of
the workshop. Presentation of results on the challenge dataset
is in all cases most welcome.

Call for papers:
http://trimbot2020.webhosting.rug.nl/events/3drms/paper-submission/


**********************************
SEMANTIC RECONSTRUCTION CHALLENGE:
**********************************

Part of the workshop is a challenge on combining 3D and semantic
information in complex scenes. To this end, the dataset was rendered
from a drive through a semantically-rich virtual garden scene with many
fine structures. Virtual models of the environment will allow us to provide
exact ground truth for the 3D structure and semantics of the garden and
rendered images from virtual multi-camera rig, enabling the use of both
stereo and motion stereo information. The challenge participants will
submit their result for benchmarking in one or more categories:
the quality of the 3D reconstructions, the quality of semantic segmentation,
and the quality of semantically annotated 3D models.

More information will be available soon at:
http://trimbot2020.webhosting.rug.nl/events/3drms/challenge/


**********************************
ORGANIZERS:
**********************************
Radim Tylecek, University of Edinburgh, UK
Torsten Sattler, ETH Zurich, Switzerland
Thomas Brox, University of Freiburg, Germany
Marc Pollefeys, ETH Zurich, Switzerland / Microsoft, USA
Robert B. Fisher, University of Edinburgh, UK
Theo Gevers, University of Amsterdam, Netherlands

Contact: Radim Tylecek (rtylecek@inf.ed.ac.uk)
Add a comment...

Post has attachment
================================
Semantic Reconstruction Challenge 2018
================================
part of 3D Reconstruction meets Semantics 2018 Workshop
in conjuction with ECCV 2018 (September 9th, Munich, Germany)
http://trimbot2020.webhosting.rug.nl/events/3drms/challenge/
================================
In order to support work on questions related to the integration of 3D
reconstruction with semantics, the workshop features a semantic
reconstruction challenge. The dataset was rendered from a drive
through a semantically-rich virtual garden scene with many fine
structures. Virtual models of the environment will allow us to provide
exact ground truth for the 3D structure and semantics of the garden
and rendered images from virtual multi-camera rig, enabling the use of
both stereo and motion stereo information. The challenge participants
will submit their result for benchmarking in one or more categories:
the quality of the 3D reconstructions, the quality of semantic
segmentation, and the quality of semantically annotated 3D models.
Additionally, a dataset captured in the real garden from moving robot
is available for validation.

Given a set of images and their known camera poses, the goal of the
challenge is to create a semantically annotated 3D model of the scene.
To this end, it will be necessary to compute depth maps for the images
and then fuse them together (potentially while incorporating
information from the semantics) into a single 3D model.
Authors of the best scoring submissions will have the possibility to
present their approach and results at the workshop.

Datasets
------------
We provide the following data for the challenge:
* A synthetic training sequences consisting of
- 20k calibrated images with their camera poses,
- ground truth semantic annotations for a subset of these images,
- a semantically annotated 3D point cloud depicting the area of the
training sequence.
* A synthetic testing sequence consisting of 5k calibrated images with
their camera poses.
* A real-world validation sequence consisting of 268 calibrated images
with their camera poses.

Both training and testing data are available at
https://gitlab.inf.ed.ac.uk/3DRMS/Challenge2018.
Please see the git repository for details on the file formats.

This year we accept submissions in several categories: semantics and
geometry, either joint or separate. For example, if you have a
pipeline that first computes semantics and geometry independently and
then fuses them, we can compare how the fused result improved
accuracy.

Submission
----------------
The deadline for submitting to the challenge is July 10th (23:59 GMT).
Please follow the instructions on the website to submit your results
in the following categories:
A. Semantic Mesh
B. Geometric Mesh
C. Semantic Image Annotations
Add a comment...

Post has attachment
How robust is your vision algorithm? We are excited to announce the 1st Robust Vision Challenge (ROB) in conjunction with CVPR 2018! To assess robustness, we evaluate algorithms across various existing and novel datasets (indoors vs. outdoors, sunny vs. bad weather, various sensors). Our challenges include: (multi-view) stereo, optical flow, single image depth prediction, semantic segmentation and semantic instance segmentation. The winners of each challenge will receive $1000 (1st place) and $500 (2nd place), present at our CVPR workshop and are invited to a joint dinner after the workshop.

Visit us at: http://www.robustvision.net
Submission deadline: June 1, 2018 (no extensions)

We have now finalized all challenges and evaluation servers. All leaderboards have been populated with baseline entries. We are very much looking forward to your submissions!

The Robust Vision Challenge Team

Andreas Geiger, Matthias Nießner, Marc Pollefeys, Carsten Rother, Daniel Scharstein, Hassan Alhaija, Angela Dai, Katrin Honauer, Joel Janai, Torsten Sattler, Nick Schneider, Johannes Schönberger, Thomas Schöps, Jonas Uhrig, Jonas Wulff and Oliver Zendel
Add a comment...

Post has attachment
3DRMS - 3D Reconstruction meets Semantics - ECCV 2018 Workshop

- A Workshop, a Challenge, and peer-reviewed Papers

The Second Workshop on 3D Reconstruction Meets Semantics:
Integration of 3D Vision with Recognition and Learning

**********************************
Date: 9th September 2018 (Morning/Half-day)
Location: Munich, Germany
Website: http://trimbot2020.webhosting.rug.nl/events/3drms/
**********************************


**********************************
IMPORTANT DATES:
**********************************
Submission of full papers and extended abstracts: JULY 10TH, 2018 - 23:59 (GMT)
Submission of challenge results: JULY 10TH, 2018 - 23:59 (GMT)
Notification of acceptance: JULY 31TH, 2018
Challenge results evaluated: AUGUST 1st, 2018
Camera-ready manuscripts: LATE SEPTEMBER, 2018 (after workshop)

**********************************
INVITED SPEAKERS
**********************************
Andrew Davison, Imperial College London, UK
Thomas Funkhouser, Princeton University, USA
Christian Häne, University of California at Berkeley, USA


**********************************
SCOPE:
**********************************
The goal of this workshop is to explore and discuss new ways for
integrating techniques from 3D reconstruction with recognition and
learning. How can semantic information be used to improve the dense
matching process in 3D reconstruction techniques? How valuable is 3D
shape information for the extraction of semantic information? In the age
of deep learning, can we formulate parts of 3D reconstruction as a
learning problem and benefit from combined networks that estimate both
3D structures and their semantic labels? How do we obtain feedback-loops
between semantic segmentation and 3D techniques that improve both
components? Will this help recover more detailed 3D structures?

Topics of interest for this workshop include, but are not limited to:
* Semantic 3D reconstruction and semantic SLAM
* Learning for 3D vision
* Fusion of geometric and semantic maps
* Label transfer via 3D models
* Datasets for semantic reconstruction including synthetic dataset generation for learning
* 2D/3D scene understanding and object detection
* Joint object segmentation and depth layering
* Correspondence and label generation from semantic 3D models
* Robotics applications based on semantic reconstructions
* Semantically annotated models for augmented reality

We encourage both the submission of original work in the form of full
papers and work in progress in the form of an extended abstract. Full
papers will be peer-reviewed and will be published in the proceedings of
the workshop through Springer LNCS. Extended abstracts will not be
formally published, but we will collect the abstracts on the website of
the workshop. Presentation of results on the challenge dataset
is in all cases most welcome.

Call for papers:
http://trimbot2020.webhosting.rug.nl/events/3drms/paper-submission/


**********************************
SEMANTIC RECONSTRUCTION CHALLENGE:
**********************************

Part of the workshop is a challenge on combining 3D and semantic
information in complex scenes. To this end, the dataset was rendered
from a drive through a semantically-rich virtual garden scene with many
fine structures. Virtual models of the environment will allow us to provide
exact ground truth for the 3D structure and semantics of the garden and
rendered images from virtual multi-camera rig, enabling the use of both
stereo and motion stereo information. The challenge participants will
submit their result for benchmarking in one or more categories:
the quality of the 3D reconstructions, the quality of semantic segmentation,
and the quality of semantically annotated 3D models.

More information will be available soon at:
http://trimbot2020.webhosting.rug.nl/events/3drms/challenge/


**********************************
ORGANIZERS:
**********************************
Radim Tylecek, University of Edinburgh, UK
Torsten Sattler, ETH Zurich, Switzerland
Thomas Brox, University of Freiburg, Germany
Marc Pollefeys, ETH Zurich, Switzerland / Microsoft, USA
Robert B. Fisher, University of Edinburgh, UK
Theo Gevers, University of Amsterdam, Netherlands

Contact: Radim Tylecek (rtylecek@inf.ed.ac.uk)
Add a comment...
Wait while more posts are being loaded