Workshop at CVPR20

Invited Speakers for V4AS@CVPR’20


June 14 (Sunday). Live talks on Zoom.
We thank you all for attending the workshop. The videos of the talks will be made available soon.
Starting Time
08:00 17:00 23:00 Opening Remarks
08:10 17:10 23:10 Invited Talk: Prof. Sabine Süsstrunk (EPFL)
“Towards Augmented Vision using Invisible Radiation”
08:50 17:50 23:50 Invited Talk: Dr. Werner Ritter (Daimler AG)
“DENSE 24/7 all-weather perception system. Results from 4 years of research.”
09:30 18:30 00:30 Dark Zurich Challenge
09:30: Christos Sakaridis (ETH Zurich)
“The UIoU Dark Zurich Challenge on Uncertainty-Aware Semantic Nighttime Image Segmentation”
09:45: Qimeng Wang (Huazhong University of Science and Technology) & Nisarg A. Shah (Indian Institute Of Technology)
“Winning the UIoU Dark Zurich Challenge”
10:00 19:00 01:00 Break
10:30 19:30 01:30 Invited Talk: Prof. Qifeng Chen (HKUST)
“Sensor Data Mining for Image Enhancement and Depth Sensing from Day to Night”
11:10 20:10 02:10 Invited Talk: Prof. Jean-François Lalonde (Université Laval)
“Going beyond the clear weather assumption”
11:50 20:50 02:50 Paper Presentations
11:50: Aditya Mehta (Birla Institute of Technology and Science)
“HIDeGAN: A Hyperspectral-guided Image Dehazing GAN”
12:00: Yoshihiro Hirohashi (DENSO Corporation)
“Removal of Image Obstacles for Vehicle-mounted Surrounding Monitoring Cameras by Real-time Video Inpainting”
12:10: Rachel Blin (INSA Rouen Normandie)
“A new multimodal RGB and polarimetric image dataset for road scenes analysis”
12:20: Lei Yu (Wuhan University)
“Implicit Euler ODE Networks for Single-Image Dehazing”
12:30: Fuxun Yu (George Mason University)
“Unsupervised Domain Adaptation for Object Detection via Cross-Domain Semi-Supervised Learning”
12:35: Shuyang Dai (Duke University)
“Adaptation Across Extreme Variations using Unlabeled Bridges”
12:40: Sergey Tarasenko (Mobility Technologies Co.,Ltd)
“Streaming Network Applications for Adverse Weather and Lighting Conditions”
12:45 21:45 03:45 Break
13:20 22:20 04:20 Invited Talk: Prof. Felix Heide (Princeton University)
“Designing Cameras to Detect the “Invisible” : Computational Imaging for Adverse Conditions”
14:00 23:00 05:00 Invited Talk: Prof. Wen Li (Uni. of Electronic Science and Technology of China)
“Driving in Adverse Weather: Domain Adaptation and Beyond”
14:40 23:40 05:40 Invited Talk: Prof. Torsten Sattler (Chalmers University of Technology)
“Semantic Visual Localization in Changing Environments”
15:20 00:20 06:20 Closing Remarks

Accepted Papers

  1. Aditya Mehta, Harsh Sinha, Pratik Narang, Murari Mandal. HIDeGan: A Hyperspectral-Guided Image Dehazing GAN.
  2. Yoshihiro Hirohashi, Kenichi Narioka, Masanori Suganuma, Xing Liu, Yukimasa Tamatsu, Takayuki Okatani. Removal of Image Obstacles for Vehicle-Mounted Surrounding Monitoring Cameras by Real-Time Video Inpainting.
  3. Rachel Blin, Samia Ainouz, Stephane Canu, Fabrice Meriaudeau. A New Multimodal RGB and Polarimetric Image Dataset for Road Scenes Analysis.
  4. Jiawei Shen, Zhuoyan Li, Lei Yu, Gui-Song Xia, Wen Yang. Implicit Euler ODE Networks for Single-Image Dehazing.

Accepted Abstracts

  1. Fuxun Yu, Di Wang, Yinpeng Chen, Nikolaos Karianakis, Pei Yu , Dimitrios Lymberopoulos, Xiang Chen. Unsupervised Domain Adaptation for Object Detection via Cross-Domain Semi-Supervised Learning.
  2. Shuyang Dai, Kihyuk Sohn, Yi-Hsuan Tsai, Lawrence Carin, Manmohan Chandraker. Adaptation Across Extreme Variations using Unlabeled Bridges.
  3. Sergey Tarasenko, Fumihiko Takahashi. Streaming Network Applications for Adverse Weather and Lighting Conditions.

UIoU Dark Zurich Challenge @CVPR20

Jointly with the “Vision for All Seasons” workshop, we organize the “UIoU Dark Zurich” challenge on uncertainty-aware semantic nighttime image segmentation. The challenge uses the Dark Zurich dataset presented in the ICCV 2019 paper ” Guided Curriculum Model Adaptation and Uncertainty-Aware Evaluation for Semantic Nighttime Image Segmentation” and containing a total of 8779 images captured at nighttime, twilight, and daytime, along with the respective GPS coordinates of the camera for each image. Evaluation of semantic segmentation models on the labeled nighttime part of Dark Zurich is based on a novel, uncertainty-aware framework in which corresponding daytime images are leveraged at annotation to assign reliable semantic labels to originally indiscernible image regions beyond human recognition capability and to indeed include such invalid regions in the evaluation jointly with valid regions.