Geomagical Labs

Content

1. About the workshop
2. Important dates
3. Format and Agenda
4. Call for participation
5. Organizers
6. Program committee members

About the Workshop

Removing an existing object from a scene is a hard problem. It not only requires the completion of geometry and texture, but also the removal of associated lighting effects (e.g. shadows, interreflections).

Major industrial players (e.g. IKEA, Amazon, Apple) offer Augmented Reality (AR) applications that allow users to place furniture virtually inside a room. There is an increasing interest in also performing the opposite: virtually removing existing furniture from a room. This empowers the users to completely reimagine their homes without having to physically move a single piece of furniture.

In this workshop, we invite researchers and industry practitioners to come together and discuss possible ways to efficiently tackle the problem of object removal. We separate the problem into several key components: room layout estimation, image inpainting and shadow removal. Each component is discussed within short presentation sessions. We also invite participants to discuss any relevant work comprising, but not limited to, Diminished Reality, inverse rendering, neural radiance fields, and room geometry estimation. The last session is a live Demo of our solution, which brings these components together to form a viable commercial Furniture Eraser application.

The workshop is in conjunction with the IEEE International Symposium on Mixed and Augmented Reality (ISMAR) 2022.

Important Dates

To submit your work, please make a submission through our EasyChair portal.

NOTE: For each presented paper, the presenter/speaker needs to make at least a one-day-conference pass registration payment (expected to be $350SGD).

Format and Agenda

NOTE: We expect 10 minutes per presenter, 6-7 minutes for presentation, and 3-4 minutes for questions. These timings are subject to change as we collect and finalize the participants list.

Call for Participation

The goal of the workshop is to act as a very flexible platform for the participants to showcase their work, and open up discussions around emerging techniques in this domain. We accept submissions of length 2-6 pages excluding references. We are open to submissions of different kinds:

• ∙ Research papers.
• ∙ Position papers.
• ∙ Survey papers.
• ∙ Practical applications of existing techniques for removing objects from an indoor scene.
• ∙ Preliminary results on one or more of the relevant topics.

Submissions are invited on topics including:

• ∙ Image and/or video inpainting.
• ∙ Depth inpainting.
• ∙ Room layout estimation.
• ∙ Lighting / shadow estimation.
• ∙ Inverse Rendering.
• ∙ Shadow segmentation and/or removal.
• ∙ Object segmentation.
• ∙ Diminished reality.
• ∙ Planar reconstruction.
• ∙ Radiance fields.
• ∙ CAD approximation of an indoor scene.

The accepted papers will be published in ISMAR 2022 adjunct proceedings and IEEE Xplore. The submission template is the same as the ISMAR 2022 template for conference papers (IEEE Computer Society VGTC format).

To submit your work, please make a submission through our EasyChair portal.

Organizers

Prakhar Kulshreshtha

is an Applied Research Engineer at Geomagical Labs, Inc. He leads a research project focused on developing a commercial furniture eraser solution. He obtained his Masters in Computer Vision from Carnegie Mellon University, where he worked with Prof. Michael Kaess on long term mapping for SLAM in a dynamic environment.

His research interests are Computer Vision and Machine Learning, with a focus on utilizing Deep Learning and 3D geometry for Computer Vision applications.

Salma Jiddi

is a Research Scientist who heads the Research team at Geomagical Labs, Inc. She completed her PhD on Photometric Registration of Indoor Real Scenes using an RGB-D Camera with Application to Mixed Reality, at Technicolor and Irisa Rennes – Bretagne Atlantique, in the Rainbow team, under the supervision of Eric Marchand (Irisa) and Philippe Robert (Technicolor).

Her research interests are Computer Vision and Computer Graphics, with a focus on the geometric and photometric alignment of real and virtual worlds.

Nektarios Lianos

is an Applied Scientist, who leads the 3D Reconstruction team at Geomagical Labs, Inc. He previously obtained his Masters in Robotics at ETH Zürich, where he worked with Prof. Marc Pollefeys on visual odometry and deep learning.

His research interests include 3D Computer Vision, Machine Learning, and SLAM.

Program Committee Members

Prakhar Kulshreshtha

Salma Jiddi

Nektarios Lianos

Ajay Unagar

Brian Pugh

Luis Puig