Dengxin Dai


Welcome

Dengxin is a senior researcher at MPI for Informatics, where he heads the research group Vision for Autonomous Systems. He also works with the Computer Vision Lab at ETH Zurich on the research project TRACE-Zurich for Autonomous Driving. In 2016, he obtained his PhD at ETH Zurich under the supervision of Prof. Luc Van Gool and Prof. Gerhard Schmitt. During the PhD study, he was working on the project VarCity for City Modeling based on camera data.

His current research interests include (1) Robust Perception Algorithms in Adverse Conditions; (2) Lifelong Deep Learning (Domain Adaptation, Semi-Supervised Learning, Self-Supervised Learning, Weakly-Supervised Learning, Active Learning ...); (3) Sensor Fusion; (4) Multi-Task Learning; (5) Auditory Perception; and (6) Map-based Perception.

News

2021/07/27: I am starting a new research group Vision for Autonomous Systems at MPI for Informatics, working on deep perception for autonomous driving, especially on the scalability of deep perception methods to new domains, new tasks, and to new sensors. I am hiring PhD students!

2021/07/27: Six papers have been accepted to ICCV 2021.

2021/04/29: 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 here.

2021/04/20: We will organize DeepMTL: Workshop on Multi-Task Learning in Computer Vision, along with ICCV 2021. We have an excellent lineup of speakers and will accept full-length papers.

2021/03/01: We will organize the 3rd workshop Vision for All Seasons: Adverse Weather and Lighting Conditions, along with CVPR 2021. We have an excellent lineup of speakers.

Publications

2021

"Learnable Online Graph Representations for 3D Multi-Object Tracking", Jan-Nico Zaech, Dengxin Dai, Alex Liniger, Martin Danelljan, Luc Van Gool, 2021.
[Paper] [BibTex]

"Hyperspectral Image Super-Resolution with Spectral Mixup and Heterogeneous Datasets", Ke Li, Dengxin Dai, Ender Konukoglu, Luc Van Gool, 2021.
[Paper] [Code] [BibTex]

"End-to-End Urban Driving by Imitating a Reinforcement Learning Coach", Zhejun Zhang, Alexander Liniger, Dengxin Dai, Fisher Yu, Luc Van Gool, ICCV 2021.
[Paper] [Code] [BibTex]

"Fog Simulation on Real LiDAR Point Clouds for 3D Object Detection in Adverse Weather", Martin Hahner, Christos Sakaridis, Dengxin Dai, Luc Van Gool, ICCV 2021.
[Paper] [Code] [BibTex]

"ACDC: The Adverse Conditions Dataset with Correspondences for Semantic Driving Scene Understanding", Christos Sakaridis, Dengxin Dai, Luc Van Gool, ICCV 2021.
[Paper] [Data & Benchmark] [BibTex]

"Domain Adaptive Semantic Segmentation with Self-Supervised Depth Estimation", Qin Wang, Dengxin Dai, Lukas Hoyer, Olga Fink, Luc Van Gool, ICCV 2021.
[Paper] [Code] [BibTex]

"mDALU: Multi-Source Domain Adaptation and Label Unification with Partial Datasets", Rui Gong, Dengxin Dai, Yuhua Chen, Wen Li, Luc Van Gool, ICCV 2021.
[Paper] [BibTex]

"DLOW: Domain Flow and Applications", Rui Gong, Wen Li, Yuhua Chen, Dengxin Dai, Luc Van Gool, IJCV 2021.
[Paper] [Code] [BibTex]

"Three Ways to Improve Semantic Segmentation with Self-Supervised Depth Estimation", Lukas Hoyer, Dengxin Dai, Yuhua Chen, Adrian Köring, Suman Saha, Luc Van Gool, CVPR 2021.
[Paper] [Code] [Video] [BibTex]

"Self-Aligned Video Deraining with Transmission-Depth Consistency", Wending Yan, Wenhang Yang, Robby T. Tan, and Dengxin Dai. CVPR 2021.
[Paper] [Supp] [BibTex]

"Cluster, Split, Fuse, and Update: Meta-Learning for Open Compound Domain Adaptive Semantic Segmentation", Rui Gong, Yuhua Chen, Danda Pani Paudel, Yawei Li, Ajad Chhatkuli, Wen Li, Dengxin Dai and Luc Van Gool, CVPR 2021.
[Paper] [BibTex]

"Spectral Tensor Train Parameterization of Deep Learning Layers", Anton Obukhov, Maxim Rakhuba, Alex Liniger, Zhiwu Huang, Stamatios Georgoulis, Dengxin Dai and Luc Van Gool, AISTATS, 2021.
[Paper] [Project] [Code] [Video] [BibTex]

"Multi-Task Learning for Dense Prediction Tasks: A Survey", Simon Vandenhende, Stamatios Georgoulis, Marc Proesmans, Dengxin Dai and Luc Van Gool, PAMI 2021.
[arXiv] [BibTex] [Code]

"Scale-Aware Domain Adaptive Faster R-CNN", Yuhua Chen, Haoran Wang, Wen Li, Christos Sakaridis, Dengxin Dai, Luc Van Gool, IJCV 2021.
[Paper] [BibTex] [Code]

"Map-Guided Curriculum Domain Adaptation and Uncertainty-Aware Evaluation for Semantic Nighttime Image Segmentation", Christos Sakaridis , Dengxin Dai and Luc Van Gool, PAMI, 2021.
[Paper] [BibTex] [Project] [Code]

"Analogical Image Translation for Fog Generation", Rui Gong, Dengxin Dai, Yuhua Chen, Wen Li, Luc Van Gool, AAAI, 2021.
[Paper] [BibTex]

"Improving Point Cloud Semantic Segmentation by Learning 3D Object Detection", Ozan Unal, Luc Van Gool, and Dengxin Dai, WACV, 2021.
[Paper] [BibTex] [Project]

2020

"Talk2Nav: Long-Range Vision-and-Language Navigation with Dual Attention and Spatial Memory", Arun Balajee Vasudevan, Dengxin Dai and Luc Van Gool, IJCV, 2020.
[Paper] [BibTex] [Project] [Video]

"T-Basis: a Compact Representation for Neural Networks", Anton Obukhov, Maxim Rakhuba, Stamatios Georgoulis, Menelaos Kanakis, Dengxin Dai and Luc Van Gool, ICML, 2020.
[Paper] [BibTex] [Code] [Video]

"Semantic Object Prediction and Spatial Sound Super-resolution with Binaural Sounds", Arun Balajee Vasudevan, Dengxin Dai and Luc Van Gool, ECCV (spotlight) 2020.
[arXiv] [BibTex] [Project] [Data] [Code]

"Weakly Supervised 3D Object Detection from Lidar Point Cloud", Qinghao Meng, Wenguan Wang, Tianfei Zhou, Jianbing Shen, Luc Van Gool, and Dengxin Dai, ECCV 2020.
[arXiv] [BibTex] [Code]

"Off-Policy Reinforcement Learning for Efficientand Effective GAN Architecture Search", Yuan Tian, Qin Wang, Zhiwu Huang, Wen Li, Dengxin Dai, Minghao Yang, Jun Wang, and Olga Fink, ECCV 2020.
[Paper] [BibTex] [Code] [Video]

"Nighttime Defogging Using High-Low Frequency Decomposition and Grayscale-Color Networks", Wending Yan, Robby T. Tan, and Dengxin Dai, ECCV 2020.
[Paper] [BibTex]

"Don’t Forget The Past: Recurrent Depth Estimation from Monocular Video", Vaishakh Patil, Wouter Van Gansbeke, Dengxin Dai and Luc Van Gool, IEEE Robotics and Automation Letters (RA-L), 2020.
[arXiv] [BibTex]

"Depth Estimation from Monocular Images and Sparse Radar Data", Juan-Ting Lin, Dengxin Dai, and Luc Van Gool, IROS 2020.
[Paper] [BibTex] [Code]

"Action Sequence Predictions of Vehicles in Urban Environments using Map and Social Context", Jan-Nico Zaech, Dengxin Dai, Alex Liniger, and Luc Van Gool, IROS 2020.
[Paper] [BibTex]

"Learning Accurate and Human-Like Driving using Semantic Maps and Attention", Simon Hecker, Dengxin Dai, Alex Liniger, and Luc Van Gool, IROS, 2020.
[Paper] [BibTex] [Project] [Data] [Workshop]

2019

"Guided Curriculum Model Adaptation and Uncertainty-Aware Evaluation for Semantic Nighttime Image Segmentation", Christos Sakaridis , Dengxin Dai and Luc Van Gool, ICCV 2019.
[arXiv] [BibTex] [Project] [Code]
A new method for learning, data annotation, method evaluation and a new dataset for semantic nighttime image segmentation.

"Semantic Understanding of Foggy Scenes with Purely Synthetic Data", Martin Hahner, Dengxin Dai, Christos Sakaridis, Jan-Nico Zaech , Luc Van Gool, ITSC 2019.
[PDF] [BibTex] [Project] [Code]

"Learning a Curve Guardian for Motorcycles", Simon Hecker, Alex Liniger, Henrik Maurenbrecher, Dengxin Dai and Luc Van Gool, ITSC 2019.
[Paper] [BibTex] [Venturebeat] [MOTO.IT] [Rideapart] ...
A new road curvature warning system for motorcycles, combining advances in computer vision, optimal control and mapping.

"Texture Underfitting for Domain Adaptation", Jan-Nico Zaech, Martin Hahner, Dengxin Dai, Luc Van Gool, ITSC 2019.
[PDF] [BibTex] [Project]

"Real-time 3D Traffic Cone Detection for Autonomous Driving",
Ankit Dhall , Dengxin Dai and Luc Van Gool, IEEE Intelligent Vehicles Symposium (IV), 2019.
[Paper] [BibTex] [Video] [Project]
A new method for real-time 3D traffic cone detection, deployed in a real racing car (see our video demo)

"Curriculum Model Adaptation with Synthetic and Real Data for Semantic Foggy Scene Understanding", Dengxin Dai, Christos Sakaridis, Simon Hecker and Luc Van Gool, International Journal of Computer Vision (IJCV), 2019.
[Paper] [Arxiv] [BibTex] [Project] [Code] [Data]
A novel curriculum model adaptation method to transfer semantic models from clear-weather condition to foggy weather condition by combining synthetic and real data. Synthetic and real fog datasets are provided.

2018

"End-to-End Learning of Driving Models with Surround-View Cameras and Route Planners", Simon Hecker, Dengxin Dai, Luc Van Gool, ECCV 2018.
[PDF] [Poster] [arXiv] [BibTex] [Demo Video] [Project]

"Failure Prediction for Autonomous Driving", Simon Hecker, Dengxin Dai, Luc Van Gool, IEEE Intelligent Vehicles Symposium, 2018.
[PDF] [BibTex]

"Model Adaptation with Synthetic and Real Data for Semantic Dense Foggy Scene Understanding", Christos Sakaridis, Dengxin Dai, Simon Hecker, Luc Van Gool, ECCV 2018.
[arXiv] [PDF] [BibTex] [Project] [Code] [Data]

"Semantic Foggy Scene Understanding with Synthetic Data", Christos Sakaridis, Dengxin Dai, Luc Van Gool, International Journal of Computer Vision, 2018.
[PDF (preprint)] [PDF (published)] [BibTex] [Project] [Data] [Code]

"Dark Model Adaptation: Semantic Image Segmentation from Daytime to Nighttime", Dengxin Dai, Luc Van Gool, International Conference on Intelligent Transportation Systems (ITSC), 2018.
[PDF] [BibTex] [Project] [Data]

"Domain Adaptive Faster R-CNN for Object Detection in the Wild", Yuhua Chen, Wen Li, Christos Sakaridis, Dengxin Dai, Luc Van Gool CVPR 2018.
[PDF] [BibTex] [Code]

"Object Referring in Videos with Language and Human Gaze", Arun Balajee Vasudevan, Dengxin Dai, Luc Van Gool. CVPR 2018.
[PDF] [BibTex] [Demo Video ] [Project]

"Object Referring in Visual Scene with Spoken Language", Arun Balajee Vasudevan, Dengxin Dai, Luc Van Gool, WACV 2018.
[PDF] [BibTex] [Demo Video ] [Project]

2017

"PathTrack: Fast Trajectory Annotation with Path Supervision", S. Manen, M. Gygli, D. Dai, L. Van Gool, ICCV 2017.
[paper] [BibTex] [Data] [Demo Video] [Project]

"Deep Domain Adaptation by Geodesic Distance Minimization", Yifei Wang, Wen Li, Dengxin Dai, Luc Van Gool, ICCV Workshop, 2017.
[paper] [BibTex]

"VarCity - The Video: the Struggles and Triumphs of Leveraging Fundamental Research Results in a Graphics Video Production", K. Vanhoey, C. Porto de Oliveira, H. Riemenschneider, A. Bodis-Szomoru, S. Manen, D. Paudel, M. Gygli, N. Kobyshev, T. Kroeger, D. Dai, L. Van Gool,, Siggraph Talk 2017.
[paper] [BibTex] [Video] [Project]

"Domain Generalization and Adaptation using Low Rank Exemplar SVMs", W. Li, Z. Xu, D. Xu, D. Dai, and L. Van Gool , T-PAMI.
[paper] [BibTex]

"Speech-based Visual Question Answering", Ted Zhang, Dengxin Dai, Tinne Tuytelaars, Marie-Francine Moens, Luc Van Gool
[paper] [BibTex] [Code] [Data]

2016

"Fast Optical Flow using Dense Inverse Search", T. Kroeger, R. Timofte, D. Dai, and L. Van Gool , ECCV 2016.
[paper] [project] [code] [BibTex]

"Scale-Aware Alignment of Hierarchical Image Segmentation", Y. Chen, D. Dai, J. Pont-Tuset and L. Van Gool , CVPR 2016.
[paper] [code] [BibTex]

"Fast Algorithms for Linear and Kernel SVM+", W. Li, D. Dai, M. Tan, D. Xu and L. Van Gool , CVPR 2016.
[paper] [code ] [BibTex]

"Is Image Super-resolution Helpful for Other Vision Tasks?", D. Dai, Y. Wang, Y. Chen, L. Van Gool, WACV 2016.
[paper] [ arXiv link] [project ] [ code ] [BibTex]

"Leveraging single for multi-target tracking using a novel trajectory overlap affinity measure", S. Manen, R. Timofte, D. Dai, L. Van Gool, WACV 2016.
[paper] [Supp. Material] [BibTex]

"Unsupervised High-level Feature Learning by Ensemble Projection for Semi-supervised Image Classification and Image Clustering", Dengxin Dai and Luc Van Gool, Tech. Report.
[paper]     [project]     [code]     [BibTex]    

2015

"Metric imitation by manifold transfer for efficient vision applications", D. Dai, T. Kroeger, R. Timofte, L. Van Gool, CVPR 2015.
[paper] [poster] [supp. mat.] [abstract] [project] [code] [BibTex]

"Joint Vanishing Point Extraction and Tracking", T. Kroeger, D. Dai, L. Van Gool, CVPR 2015, (Oral).
[paper] [supp. mat.] [abstract] [project] [BibTex]

"Jointly Optimized Regressors for Image Super-resolution", D. Dai, R. Timofte, L. Van Gool, Eurographics 2015, (Oral).
[paper] [project] [code] [slide] [BibTex]

"Discovery of Sets of Mutually Orthogonal Vanishing Points in Videos", T. Kroeger, D. Dai, R. Timofte, L. Van Gool, WACV-W 2015.
[paper] [BibTex]

2014

"The Synthesizability of Texture Examples", Dengxin Dai, Hayko Riemenschneider, and Luc Van Gool, CVPR 2014.
[project] [code] [data] [paper] [BibTex]

"Latent Dictionary Learning for Sparse Representation based Classification", Meng Yang, Dengxin Dai, Linlin Shen, and Luc Van Gool, CVPR 2014.
[paper] [BibTex]

2013

"Example-based Facade Texture Synthesis", Dengxin Dai, Hayko Riemenschneider, Gerhard Schmitt, and Luc Van Gool, ICCV 2013 .
[project]     [data]     [PDF]     [BibTex]    

"Ensemble Projection for Semi-supervised Image Classification", Dengxin Dai and Luc Van Gool, ICCV 2013 .
[project]     [code]     [PDF]     [BibTex]    

2012

"Learning Domain Knowledge for Facade Labeling", Dengxin Dai, Mukta Prasad, Gerhard Schmitt, and Luc Van Gool, ECCV 2012 .
[project]     [PDF]     [BibTex]    

"Ensemble Partitioning for Unsupervised Image Categorization", Dengxin Dai, Mukta Prasad, Christian Leistner, and Luc Van Gool, ECCV 2012 .
[project]     [PDF]     [BibTex]    

"SAR-Based Terrain Classification Using Weakly Supervised Hierarchical Markov Aspect Models", Wen Yang, Dengxin Dai, Bill Triggs, and Gui-Song Xia, IEEE Trans. Image Process , 2012.
[PDF]     [BibTex]    

Previous

"Multilevel Local Pattern Histogram for SAR Image Classification", Dengxin Dai, Wen Yang, and Hong Sun, IEEE Geosci. Remote Sens. Lett. , 2011.
[code]     [PDF]     [BibTex]    

"Satellite Image Classification via Two-Layer Sparse Coding With Biased Image Representation ", Dengxin Dai and Wen Yang IEEE Geosci. Remote Sens. Lett. , 2011.
[data]     [PDF]     [BibTex]    

"Discovering Scene Categories by Information Projection and Cluster Sampling", Dengxin Dai, Tianfu Wu, and Song-Chun Zhu CVPR 2010
[PDF]     [BibTex]    

Teaching

  • Lead instructor: Deep Learning for Autonomous Driving at ETH Zurich in 2020 and 2021.
  • Teaching assistant: Image Analysis and Computer Vision at ETH Zurich in 2012-2014.
  • Academic Service

  • Co-organizer of Workshop DeepMTL: Deep Multi-Task Learning at ICCV'21.
  • Lead Guest Editor of the Special Issue "Computer Vision for All Seasons" of IJCV.
  • Area Chair of WACV 2020 and CVPR 2021.
  • Senior Program Committee Member: AAAI 2020 and IJCAI 2019.
  • Lead workshop organizer: Vision for All Seasons at CVPR'19, CVPR'20, and CVPR'21.
  • Lead workshop organizer: Autonomous Driving workshop at ICCV'19.
  • Co-organizer of challenge Learning to Drive at ICCV'19
  • Co-organizer of challenge Nighttime Semantic Image Segmentation at CVPR'20.
  • Jury member for the Pioneer Fellowship Program at ETH Zurich.
  • External Examiner of Doctoral Exams: EPFL and NUS
  • Regular Reviewers: CVPR, ICCV, ECCV, NeurIPS, ICLR, ICML, ICRA, IROS, AAAI, IJCV and PAMI.
  • Code

  • The code of some of my work can be found at the corresponding project page!
  • Bing Image Crawler : A Shell + Python script to download images from Bing image search!
  • A matlab wrapper (linux) for the Ranking SVM !
  • Contact Me

    Computer Vision Laboratory
    Sternwartstrasse 7
    ETH Zentrum
    CH - 8092 Zürich, Switzerland
    Tel: +41 44 63 25426
    Fax: +41 44 63 21199
    Office: ETF C112
    E-mail: dai[at]vision.ee.ethz.ch