TUM Computer Vision Group

Startseite
Deutschland
Garching Bei Munich
TUM Computer Vision Group

Welcome to the TUM Computer Vision Group! http://vision.in.tum.de

28/03/2022

The code for DM-VIO is now published: https://github.com/lukasvst/dm-vio

Also, we are excited to share that the paper will be presented at ICRA this year.

Source code for the paper DM-VIO: Delayed Marginalization Visual-Inertial Odometry - GitHub - lukasvst/dm-vio: Source code for the paper DM-VIO: Delayed Marginalization Visual-Inertial Odometry

12/01/2022

We are excited to announce that our new work "DM-VIO: Delayed Marginalization Visual-Inertial Odometry" has been accepted at RA-L. It is a new VIO system with state-of-the art performance on three datasets.
The code for our approach will be published soon.
Paper:https://arxiv.org/pdf/2201.04114.pdf
Code coming soon: https://github.com/lukasvst/dm-vio
Youtube: https://youtu.be/7iep3BvcJPU
Project Page: http://vision.in.tum.de/dm-vio

Published at IEEE Robotics and Automation Letters.Project Page: http://vision.in.tum.de/dm-vioCode coming soon at: https://github.com/lukasvst/dm-vio Authors...

24/12/2021

🎄Merry Christmas!🎄
The code of TANDEM (CoRL 2021 & 🏆Best Demo Award🏆 3DV 2021) is published. Please check out!
Code: https://github.com/tum-vision/tandem
Project: https://vision.in.tum.de/research/vslam/tandem
Video: https://youtu.be/L4C8Q6Gvl1w

Conference on Robot Learning (CoRL) 2021 poster & demo.GitHub:https://github.com/tum-vision/tandemarXiv:https://arxiv.org/abs/2111.07418OpenReview:https://op...

06/12/2021

We are happy to share Daniel Cremers' TEDx talk on 3D computer vision:

Teaching computers to understand the world from cameras has incredible potential for all kinds of applications: self-driving cars, medical image analysis, au...

13/10/2021

We are happy to present our paper "Square Root Marginalization for Sliding-Window Bundle Adjustment" at ICCV this week.

In this work we demonstrate how square-root estimation techniques can make optimization-based sliding-window odometry faster and numerically more stable. We are about to release the implementation as an update to the VO / VIO of Basalt, which now runs with single-precision floats without loss of accuracy and faster than ever (we measure up to 10x real-time on a multi-core desktop for datasets like EuRoC or KITTI).

Join us live at ICCV in Session 10A & 10B: Wednesday Oct 13, 5pm EDT / 11pm CEST and Friday Oct 15, 10am EDT / 4pm CEST.

Video: https://youtu.be/5JCwMSZr4IM
Paper:https://arxiv.org/abs/2109.02182.pdf
Code: https://gitlab.com/VladyslavUsenko/basalt/-/issues/53
Project Page: https://go.vision.in.tum.de/rootvo

This is the ICCV 2021 presentation video for our work:Square Root Marginalization for Sliding-Window Bundle AdjustmentAuthors: Nikolaus Demmel, David Schuber...

11/08/2021

We are excited to announce that the code (training + pretrained models) of MonoRec (CVPR 2021) is released on Github! Check it out!

Code: https://github.com/Brummi/MonoRec
Paper: https://arxiv.org/abs/2011.11814
Video: https://youtu.be/-gDSBIm0vgk
Presentation: https://youtu.be/XimdlXUamo0

CVPR 2021Publication:MonoRec: Semi-Supervised Dense Reconstruction in Dynamic Environments from a Single Moving CameraAuthors:Felix Wimbauer, Nan Yang, Lukas...

23/06/2021

[CVPR 2021] In CVPR 2021, we are presenting "i3DMM: Deep Implicit 3D Morphable Model of Human Heads", joint work by Tarun Yenamandra, Ayush Tewari, Florian Bernard, Hans-Peter Seidel, Mohamed Elgharib, Daniel Cremers, and Christian Theobalt.

i3DMM is the first full head implicit representation-based 3DMM of the human heads. It is one of the first models to predict correspondences among 3D objects represented using implicits.

If you are attending CVPR this time, drop by our chat room tomorrow in paper session 10!

Code: https://github.com/tarun738/i3DMM
Paper: https://openaccess.thecvf.com/content/CVPR2021/html/Yenamandra_i3DMM_Deep_Implicit_3D_Morphable_Model_of_Human_Heads_CVPR_2021_paper.html
Video: https://youtu.be/4pYzV3ButPY

We present the first deep implicit 3D morphable model (i3DMM) of full heads. Unlike earlier morphable face models it not only captures identity-specific geom...

22/06/2021

[CVPR 2021] RootBA: Square Root Bundle Adjustment, joint work by Nikolaus Demmel, Christiane Sommer, Daniel Cremers and Vladyslav Usenko will be presented at CVPR this week.

In RootBA we propose null-space marginalization as an alternative to the conventionally used Schur Complement trick to solve large Bundle Adjustment problems more efficiently and with improved numerical stability.

The CVPR live Q&A is during Paper Session 9, Thursday, June 24, 6:00 – 8:30 am EDT (12:00 - 14:30 CEST)

Video: https://youtu.be/kAhmjNL8B-U
Code: https://github.com/NikolausDemmel/rootba
Paper: https://arxiv.org/abs/2103.01843
Project Page: https://go.vision.in.tum.de/rootba

This is the CVPR 2021 presentation video for our work:Square Root Bundle Adjustment for Large-Scale ReconstructionAuthors: Nikolaus Demmel, Christiane Sommer...

21/06/2021

[CVPR 2021] MonoRec: Semi-Supervised Dense Reconstruction in Dynamic Environments from a Single Moving Camera, joint work by Felix Wimbauer, Nan Yang, Lukas von Stumberg, Niclas Zeller, and Daniel Cremers, will be present at CVPR 2021 this week.
MonoRec predicts the dynamic object masks, dense depth of the static background, and dense depth of the dynamic objects in one pass from a moving camera.

The CVPR live Q&A session is Tuesday, June 22, 2021, 22:00 – 24:30 EDT.

Please find more details at
Demo Video: https://youtu.be/-gDSBIm0vgk
Presentation Video: https://youtu.be/XimdlXUamo0
Code: https://github.com/Brummi/MonoRec
Paper: https://arxiv.org/abs/2011.11814
Project: https://vision.in.tum.de/monorec

CVPR 2021Publication:MonoRec: Semi-Supervised Dense Reconstruction in Dynamic Environments from a Single Moving CameraAuthors:Felix Wimbauer, Nan Yang, Lukas...

04/06/2021

Next Thursday, 10.06, at 3pm, we host Bernt Schiele from Max Planck Institute for Informatics in our TUM AI Lecture series.
He will be giving a talk on “Neuroexplict Modeling for more Interpretability and Performance in Computer Vision”.

Join us live on Youtube!
https://youtu.be/jZudw5eP1as

Looking forward to your participation!

Title: Neuroexplict Modeling for more Interpretability and Performance in Computer VisionAbstract: Computer Vision has been revolutionized by Machine Learnin...

Adresse

Department Of Computer Science Informatik 9 Boltzmannstrasse 3
Garching Bei Munich
85748

Telefon

+498928917756

Webseite

http://vision.in.tum.de/

Benachrichtigungen

Lassen Sie sich von uns eine E-Mail senden und seien Sie der erste der Neuigkeiten und Aktionen von TUM Computer Vision Group erfährt. Ihre E-Mail-Adresse wird nicht für andere Zwecke verwendet und Sie können sich jederzeit abmelden.