It is with great pleasure that we announce that MIDL 2022 will be hosted in Zürich! We are looking forward to welcoming everyone to the charming city as well as stimulating a fruitful exchange of research.
MIDL :
MIDL 2022:
Lukas Hoyer, PhD student at CVL, has been honored with the ETH Medal for Outstanding Master's Theses.
Lukas conducted his Master's project on "Improving Semantic Segmentation with Self-Supervised Depth Estimation" advised by Dengxin Dai and Luc Van Gool.
Sergi Caelles (
@skprat
) has successfully defended his PhD thesis with the title "Video
Object Segmentation: Methods and Datasets" via video conference. Congratulations and
good luck with your future endeavours!
Glad to share that our paper "Constrained Optimization to Train Neural Networks on Critical and Under-Represented Classes" by S. Sangalli, E. Erdil, A. Hoetker, O. Donati and E. Konukoglu got accepted for publication at NeurIPS 2021.
ACDC Dataset Released
ACDC is a new large-scale driving dataset for training and testing semantic segmentation algorithms on adverse visual conditions, such as fog, nighttime, rain, and snow.
The dataset and associated benchmarks
are now available at
Christos Sakaridis has successfully defended his PhD
His PhD thesis is titled "Semantic Understanding of Driving Scenes in Adverse Conditions". It introduces methods and datasets for improving the performance of semantic scene understanding algorithms under adverse conditions.
Fabian Mentzer (
@mentzer_f
) has successfully defended his PhD. In his thesis, he presented four algorithms for neural image compression, two for lossless compression, and two for lossy compression. To learn more about the underlying publications, see:
We're very excited about the four papers that will be presented from CVL at MICCAI this week. Below is a small teaser about work on reconstruction, shape and distribution modeling as well as reinforcement learning applied to medical imaging challenges.
Fabien Péan has successfully defended his PhD thesis entitled "Computational Modeling and Simulation of the Shoulder for Functional Analysis of Orthopedic Surgery Outcomes".
We are excited to share our
#NeurIPS2020
papers presented this week from our lab. Check the below teaser to get a glimpse of the topics. See you at the poster sessions!
Simon Hecker has successfully completed his doctoral thesis on the subject of autonomous driving with the title: “Vision and Data Driven Methods for Automotive Challenges: From end-to-end driving to curve warning systems”
Glad to share one of the accepted papers at CVPR2021:
- Learning to Relate Depth and Semantics for Unsupervised Domain Adaptation -
Suman Saha, Anton Obukhov, Danda Pani Paudel, Menelaos Kanakis, Yuhua Chen, Stamatios Georgoulis, Luc Van Gool
MS projects titled "Neural Architecture Search for Efficient Uncalibrated Deep Photometric Stereo" and "Neural Radiance Fields Approach to Deep Multi-View Photometric Stereo" accepted for publication at IEEE WACV'22. The projects were advised by S. Kumar, V.
Ferrari, and L.V Gool
Paper by Dr. Richard Rau et al. is runner-up for Best Paper Award in the Elsevier Medical Image Analysis journal MICCAI Special Issue. The authors from the CAiM group led by Prof. Goksel propose a new ultrasound imaging biomarker, e.g. for breast cancer.
Dr. Dengxin Dai made it into the Communications of the ACM. He was asked how he and his team let computer vision systems work even under the
worst of weather conditions, thick fog... A real challenge for autonomous cars !
More at
The Computer Vision Lab at ETH is heavily involved in the application of the new AI and of deep learning. ETH in turn is a proud member of ELLIS, a Europe-wide network of leading organizations in the area of AI and DL.
Check out (closing Dec 1, 2020):
A Real-Time Online Learning Framework for Joint 3D Reconstruction and Semantic Segmentation of Indoor Scenes." is accepted for publication at IEEE, RAL 22. Mr. Davide Menini carried out the project under the supervision of S.Kumar and L.V Gool in collaboration with Google Zurich
New paper published by
Tobias Fischer(
@TobiasFischer11
)
and
Dr. Fisher Yu (
@DrFisherYu
)
in collaboration with UC Berkeley, the National Tsing Hua University in Taiwan : "Monocular Quasi-Dense 3D Object Tracking"
We are very proud to announce that Dr. Dengxin Dai has received the Golden Owl Award 2021 of ETH Zürich for exceptional teaching. The Owl is awarded by the ETH student association VSETH.
Link:
Computer Vision Lab (CVL) made substantial contributions to the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2020.
This video highlights several CVL papers published with CVPR 2020.
Our paper "Efficient Conditional GAN Transfer with Knowledge Propagation across Classes" is accepted to
#CVPR2021
.
A collaboration with Z. Huang, D. P. Paudel, A. Chhatkuli, and Prof. Luc Van Gool
@CVL_ETH
.
Preprint:
Code (soon):
We are happy to announce the new paper published:
"Normative Ascent with Local Gaussians
for Unsupervised Lesion Detection".
Xiaoran Chen, Nick Pawlowski, Ben Glocker,
Ender Konukoglu
The new work of CVL research group won the Champion in the weakly supervised semantic segmentation track of CVPR2020 LID Challenge. The corresponding workshop paper "Mining Cross-Image Semantics for Weakly Supervised Semantic Segmentation" won the best paper prize.
Simon Hecker has been awarded with a ETH Pioneer Fellowship. His project “Aegis Rider Technology” will develop new safety systems for powered two-wheeled (PTW) vehicles such as motorcycles, e-bikes and e-scooters.
"Uncalibrated Neural Inverse Rendering for Photometric Stereo of General Surfaces"
by Berk Kaya, Suryansh Kumar, Carlos Oliveira, Vittorio Ferrari, and Luc van Gool in collaboration with Google Research is accepted at CVPR 2021.
Check out our CVPR 2022 work "Uncertainty Aware Deep Multi-View Photometric Stereo".
The project was completed at CVL, ETH Zurich, in collaboration with Google Research Zurich.
arXiv: ,
Youtube: ,
#cvpr2022
BMIC goes to ICLR 2023 in Rwanda! Congratulations to Tianfei Zhou and Gustav Bredell for their accepted papers. The links to the papers can be found below.
Bhaskara Rao Chintada has successfully defended his PhD.
As part of his thesis work, he developed new imaging methods as additional biomarkers to complement existing ultrasound imaging modalities, shear wave elasticity imaging (SWEI) and ultrasound computed tomography (USCT).
Happy to see that our approach OGR3MOT ranked second in the nuScenes Tracking Challenge at ICRA2021.
It uses a graph structure to jointly represent detection and track states and processes information with a fully trainable Neural Message Passing network.
Scene understanding from visual data is a highly demanding research topic. At the Migros Lunch & Learn Event Suman Saha presented the idea of core vision-based problems, concluding with promising qualitative results of methods established at CVL.
The new work of Zhicheng Fang on monocular depth estimation will be presented
@wacv2020
! This comprehensive study can help researchers on depth estimation model design and training. Code:
The ACDC Challenge 2022 on semantic segmentation in adverse visual conditions is organized in conjunction with the CVPR 2022 workshop Vision for All Seasons. The challenge is based on the public ACDC benchmarks and it will run until June 3, 2022.
Segmentation can be a difficult task on medical images. To not suppress this uncertainty Dr. Baumgartner proposed a probabilistic hierarchical model that allows to sample from the distribution of segmentations
@miccai2019
Paper:
The new work of CVL research group on face anti-spoofing will be presented
@CVPR2020
Biometric workshop! This novel technique can help to address the inherent "domain shift" problem present in face anti-spoofing.
Dr. Sakaridis will lead an independent research project with the support of the Career Seed Award in 2023, titled "Nighttime Photorealistic Simulation for Robust Semantic Driving Scene Understanding
Autonomous driving will revolutionize our understanding of
transportation. Within the TRACE project, CVL is advancing algorithms for perception and image understanding to enable safe autonomous vehicles. Visit us on: .
The new work of CVL research group "Domain Agnostic Feature Learning for Face Anti-spoofing" received the "Best Paper Award" at the CVPR 2020 Biometrics Workshop. Oral Video: ; Paper: .
Neural Architecture Search of SPD Manifold Networks" completed by Rhea Sukthanker at CVL Lab under the supervision of Zhiwu Huang, Suryansh Kumar, and Luc van Gool got accepted for publication at IJCAI 2021."
arXiv Link:
Link:
NomadZ held an invited talk at this year's WESYP conference, the IEEE Western European Students and Young Professionals congress. It took place on September 15-17 at EPFL.
Giuliano Albanese and Zichong Li explained RoboCup and presented the current progress of NomadZ.
The Aegis Rider project was introduced in a live session at the Swiss Digital Days TV on 03.11.2020. To watch the full video click here (German):
To watch the Aegis Rider project click here:
About Aegis:
CVL joins the:
Nachtaktiv Hop-On on 20.10.2022
"Augen für's Auto"
In a live demo it is shown how multiple sensors are used in combination to enable the car to see its surroundings.
For more information please visit the Website:
NomadZ took part in the Swiss Robotics Day held in Zürich on November 3rd, where we played autonomous robot soccer. The event was a hub of fascinating innovation, featuring presentations, projects, and leading companies in the field of robotics.