Omnidirectional Computer Vision

2nd Workshop

in conjunction with IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'2021)

Accepted papers will be published in IEEE Xplore!


Friday, June 25, 2021 (virtually)

9:00am- 5:00pm US Central Time


WATCH on YouTube Live

INTERACT with authors on Discord

(jump to the program)



Click here for the 2020 edition of the workshop

Motivation

From immersive experiences to autonomous driving to medical imaging and more, we are increasingly seeing that maximizing a camera’s field of view can solve real-world problems. In the past few years, omnidirectional imaging has grown in interest, driven in part by the desire to maximize the amount of content and context encapsulated by a single image. Fisheye cameras installed in modern vehicles and commodity omnidirectional cameras from companies like Ricoh and Insta360 have helped to increase the popularity of this imaging modality and have opened the door for more consumer-facing applications in the automotive, home services and real estate industries, among others. Our workshop seeks to provide a link between the formative research that supports these advances and the realization of commercial products that leverage this technology. We want to encourage the development of new algorithms and applications for this imaging modality that will continue to drive this engine of progress.

Ethical Considerations

Wide field of view and omnidirectional imaging require special privacy considerations as these techniques capture more expansive content than the more common perspective image. Furthermore, the common application of wide field of view imaging to surveillance and its frequent use as an "always-on" sensor for vehicles and augmented reality devices raise questions about how we can leverage the power of this imaging modality while also preserving individual privacy.

We have encouraged our speakers to address these considerations in their keynotes. We also encourage authors to explore topics around the privacy implications of omnidirectional imaging, along with privacy-preserving solutions that are unique to this modality and its applications. We further encourage authors to consider ways to leverage omnidirectional vision to address societal and environmental problems, as well as other important issues, that impact billions of people around the world.

9:00am - Welcome Remarks

9:05am - [abstract] Omnidirectional Vision "at the Edge", Dr. Michel Findeisen, 3DVisionLabs

9:30am - [abstract] A Single Piece Omnidirectional Flat Lens, Dr. Juejun Hu, MIT

9:55am - [abstract] Automotive Vision Systems - Opportunities & Challenges for the Future, Enda Ward, Valeo

10:20am - 15 MINUTE COFFEE BREAK

10:35pm - [abstract] Fast and Automatic 3D Reconstruction of Indoor Environments from Omnidirectional Images, Giovanni Pintore, CRS4

11:00am - [abstract] Scalable Capture of Omnidirectional Video, Zak Niazi, CircleOptics

11:25am - [abstract] Oral 1: 3D Object Detection from a Single Fisheye Image Without a Single Fisheye Training Image, Elad Plaut

11:35am - [abstract] Oral 2: Pano3D: A Holistic Benchmark and a Solid Baseline for 360 Depth Estimation, Georgios Albanis

11:45am - 45 MINUTE LUNCH BREAK

12:30pm - Woodscape Challenge Awards and Talks

1:15pm - [recorded talks, live discord] POSTER SESSION

2:00pm - [abstract] A Perceptual Measure for Wide Angle Lens and Camera Calibration, Dr. Jean-François Lalonde, Université Laval

2:25pm - [abstract] Casual 360° VR Photography with Motion Parallax, Dr. Christian Richardt, University of Bath

2:50pm - [abstract] Audio-Visual Floorplan Reconstruction, Senthil Purushwalkam, CMU

3:15pm - 15 MINUTE COFFEE BREAK

3:30pm - Joint Camera Geometry and Radial Distortion Estimation, Dr. Zuzana Kukelova, CTU-Prague

3:55pm - [abstract] Applications and Challenges for 360 cameras in Construction, Dr. Derek Hoiem, Reconstruct

4:20pm - [abstract] Oral 3: PanoDR: Spherical Panorama Diminished Reality for Indoor Scenes, Vasileios Gkitsas

4:30pm - [abstract] Oral 4: Fast Solvers for Minimal Radial Distortion Relative Pose Problems, Magnus Oskarsson

5:40pm - Closing Remarks

Keynote Speakers

Academia

University of Texas -Austin, Facebook AI Research

Massachusetts Institute of Technology

CTU-Prague

Université Laval

CRS4 Visual Computing Group

University of Bath

Industry

Michel Findeisen, CEO

Zak Niazi, CEO

Derek Hoiem, Chief Science Officer

Enda Ward, Camera Architecture & Technology Group Leader


Points of Contact

Marc Eder (marc.eder@yembo.ai)

Stefan Milz (website) (stefan.milz@spleenlab.ai)