6D object pose estimation

Program - August 23, 2020, UTC+1 time zone

The workshop was virtual and consisted of two Zoom sessions accessible with the ECCV delegate pass.

Session 1

10:00Opening: Tomáš Hodaň (Czech Technical University), SLIDES
10:10Invited talk 1: Stephen James (Imperial College London), From Explicit to Implicit Pose Estimation, SLIDES
10:50Invited talk 2: Shuran Song (Columbia University), Category-Level Object Pose Estimation, VIDEO
11:30BOP Challenge 2020: Tomáš Hodaň (Czech Technical University), Results & Awards, SLIDES, PAPER
12:00End of session 1

Session 2

18:00Invited talk 3: Leonidas Guibas (Stanford University), Object-Centric Situational 3D Understanding, SLIDES
18:40Invited talk 4: Dieter Fox (University of Washington, Nvidia), 6D Object Pose Estimation for Manipulation
19:20BOP Challenge 2020: Tomáš Hodaň (Czech Technical University), A Summary of Results, SLIDES, PAPER
19:30BOP Challenge 2020: Martin Sundermeyer (DLR), Photorealistic Training Images by BlenderProc
19:45BOP Challenge 2020: Yann Labbé (Inria Paris), Presentation of the Overall Best Method, SLIDES
19:50BOP Challenge 2020: Bertram Drost (MVTec), Presentation of the Best Fast Method, SLIDES
20:00End of session 2

Accepted Workshop Papers

News

Introduction

Accurate estimation of the 6D pose of an object (3D translation and 3D rotation) is of great importance to many higher-level tasks such as robotic manipulation, augmented reality, and autonomous driving.

The introduction of consumer and industrial grade RGB-D sensors and the advent of deep learning have allowed for substantial improvement in 6D object pose estimation. However, there still remain challenges to be addressed such as robustness against occlusion and clutter, scalability to multiple objects, effective synthesis of training data, and fast and reliable object modelling, including capturing of reflectance properties. Extending contemporary methods to work reliably and with sufficient execution speed in an industrial setting is still an open problem. Many recent methods focus on specific rigid objects, but pose estimation of deformable or articulated objects and of object categories is also an important research direction.

In this workshop, people working on relevant topics in both academia and industry will share up-to-date advances and identify open problems. The workshop will feature several high-quality invited talks, presentations of accepted workshop papers, and the BOP Challenge 2020 on 6D object pose estimation.

Previous workshop editions: 1st edition (ICCV 2015), 2nd edition (ECCV 2016), 3rd edition (ICCV 2017), 4th edition (ECCV 2018), 5th edition (ICCV 2019).

BOP Challenge 2020

We invite submissions to the BOP Challenge 2020 on model-based 6D object pose estimation. This year, we provide BlenderProc4BOP, an open-source and light-weight photorealistic (PBR) renderer, and a set of pre-rendered training images. This addition reduces the entry barrier of the challenge for participants working on learning-based RGB and RGB-D solutions.

Participants of the challenge have the opportunity to document their methods by submitting a short workshop paper. The paper will be reviewed by the organizational committee and accepted if it presents a method with competitive results or distinguishing features. The paper must have exactly 4 pages including references and, if accepted, will be published in the ECCV workshop proceedings.

Call for Papers

We invite submissions of long papers of unpublished work (14 pages excluding references). We also invite submissions of short papers documenting methods participating in the BOP Challenge 2020 (exactly 4 pages including references). If accepted, both long and short papers will be published in the ECCV workshop proceedings. The long papers will be also presented at the workshop. The papers must follow the format of the main conference (LaTeX template). The long papers should be submitted to the CMT system, and the short papers by e-mail to r6d.workshop@gmail.com.

The covered topics include but are not limited to:

Dates

Long-papers:
Submission deadline: July 10 July 17, 2020 (11:59PM PST)
Acceptance notification: July 31, 2020
Conference-ready version + video presentation (available at the virtual ECCV): August 11, 2020 (11:59PM PST)
Camera-ready version: September 14, 2020 (11:59PM PST)

Short-papers:
Submission deadline: August 16 August 19, 2020 (11:59PM UTC)
Acceptance notification: TBD
Camera-ready deadline: TBD

Deadline for submissions to the BOP Challenge 2020: August 16 August 19, 2020 (11:59PM UTC)
Workshop date: August 23, 2020

Sponsor

Organizers

Tomáš Hodaň, Czech Technical University in Prague, hodantom@cmp.felk.cvut.cz
Martin Sundermeyer, DLR German Aerospace Center, martin.sundermeyer@dlr.de
Rigas Kouskouridas, Scape Technologies
Tae-Kyun Kim, Imperial College London
Jiří Matas, Czech Technical University in Prague
Carsten Rother, Heidelberg University
Vincent Lepetit, ENPC ParisTech, Technical University Graz
Ales Leonardis, University of Birmingham
Krzysztof Walas, Poznan University of Technology
Carsten Steger​, Technical University of Munich, MVTec
Eric Brachmann, Heidelberg University
Bertram Drost, MVTec
Juil Sock, Imperial College London

Contact

r6d.workshop@gmail.com