The 4D Vision Workshop was held on August, 23, 2020, virtually, in conjunction with the European Conference on Computer Vision (ECCV'20) in Glasgow, Scotland.
We thank everyone, including our co-organizers, program committee members, invited speakers, panelists and authors for their contributions to this workshop! You can re-watch the workshop recording on the ECCV platform or below.
About & Program
Visual scene understanding is crucial for many practical applications in the real world. A lot of work has focused on 3D scene analysis. Considering the scene also in time, i.e. 4D visual understanding, is key for holistic understanding of the surrounding world. 3D scene analysis, 3D reconstruction, SLAM and others focus on the scene primarily without dynamic objects. At the same time video analysis has provided very advanced methods for understanding visual information in time.
Recent research e.g. in perception for autonomous driving or embodied vision, have also started to consider both aspects of scene analysis. Gleaning insights from scene understanding, video analysis, and 3D modeling and sensing in the realms of both dynamic and static scenes, we hope to shed more light on the topic of 4D vision for scene understanding.
Watch the Recording
4D Forecasting: Sequential Forecasting of 100,000 Points, Xinshuo Weng (Carnegie Mellon University), Jianren Wang (Carnegie Mellon University), Sergey Levine (UC Berkeley), Kris Kitani (Carnegie Mellon University), Nicholas Rhinehart (UC Berkeley) (paper)
PointContrast: Unsupervised Pre-training for 3D Point Cloud Understanding, Saining Xie (Facebook AI Research), Jiatao Gu (Facebook AI Research), Demi Guo (Harvard College), Charles Qi (Waymo), Leonidas Guibas (Stanford University), Or Litany (Stanford University) (paper)
Graph Neural Networks for 3D Multi-Object Tracking, Xinshuo Weng (Carnegie Mellon University), Yongxin Wang (Carnegie Mellon University), Yunze Man (Carnegie Mellon University), Kris Kitani (Carnegie Mellon University) (paper)
An Exploration of Embodied Visual Exploration, Santhosh Kumar Ramakrishnan (University of Texas at Austin), Dinesh Jayaraman (University of Pennsylvania), Kristen Grauman (Facebook AI Research & UT Austin) (paper)
Occupancy Anticipation for Efficient Exploration and Navigation, Santhosh Kumar Ramakrishnan (University of Texas at Austin), Ziad Al-Halah (UT Austin), Kristen Grauman (Facebook AI Research & UT Austin) (paper)
SoundSpaces: Audio-Visual Navigation in 3D Environments, Changan Chen (University of Texas at Austin), Unnat Jain (UIUC), Ziad Al-Halah (UT Austin), Kristen Grauman (Facebook AI Research & UT Austin) (paper)
Self-supervised Single-view 3D Reconstruction via Semantic Consistency, Xueting Li (University of California, Merced), Sifei Liu (NVIDIA), Kihwan Kim (NVIDIA), Shalini De Mello (NVIDIA Research), Varun Jampani (Google), Ming-Hsuan Yang (University of California at Merced), Jan Kautz (NVIDIA) (paper)
Topics include, but are not limited to:
3D scene understanding
Perception for autonomous driving
Ego-centric scene understanding
Sequence models for scene analysis
3D sensing of moving objects
We invite submissions of extended abstracts (up to 4 pages, including figures, and references). The submission should be in the form of a single PDF, using the ECCV conference template. Authors may optionally anonymize their submission. Submitted contributions will go through a single-blind review process.
Accepted abstracts will be presented in a poster session and selected abstracts will also feature as spotlight talks. All contributed abstracts will be made available on the workshop’s website.
Abstract Submission Deadline: 16 July, 2020.
Reviews due: 4 August, 2020.
Notification of Acceptance: 8 August, 2020.
Abstract Camera-ready Deadline: 12 August, 2020.
Workshop: Sunday 23 August, 2020
Submission site: https://cmt3.research.microsoft.com/4DV2020