Organized at ICCV 2023, October 3 (PM), Paris
October 3 (PM), 2023, UTC+2 time zone
The workshop covers topics related to estimating the 6D object pose (3D translation and 3D rotation) from RGB/RGB-D images, which is an important problem for application fields such as robotic manipulation, augmented reality and autonomous driving. The introduction of RGB-D sensors, advent of deep learning, and novel data generation pipelines led to substantial improvements in object pose estimation. Yet there remain challenges to address such as robustness against occlusion and clutter, scalability to multiple objects, effective synthetic-to-real domain transfer, fast and reliable object learning/modeling, and handling non-rigid objects and object categories. Addressing these challenges is necessary for achieving reliable solutions that can be deployed in real-world settings.
In conjunction with the workshop, we organize the BOP Challenge 2023, the fifth in a series of public competitions with the goal of capturing the status quo in the field of object pose estimation. The 2023 challenge introduces new tasks of detection, segmentation and pose estimation of objects unseen during training. By introducing this task, we wish to encourage the development of practical methods that can learn new objects on the fly just from provided 3D models, which is an important capability for industrial setups.
The workshop features four invited talks by experts in the field, presentation of the BOP Challenge 2023 results, and oral/poster presentations of accepted workshop papers and of papers invited from the main conference. The workshop is expected to be attended by people working on related topics in both academia and industry.
Previous workshop editions: 1st edition (ICCV 2015), 2nd edition (ECCV 2016), 3rd edition (ICCV 2017), 4th edition (ECCV 2018), 5th edition (ICCV 2019), 6th edition (ECCV 2020), 7th edition (ECCV 2022).
To measure the progress in the field of object pose estimation, we created the BOP benchmark and have been organizing challenges on the benchmark datasets in conjunction with the R6D workshops since 2017. This year is no exception. The BOP benchmark is far from being solved, with the pose estimation accuracy improving significantly every challenge — the state of the art moved from 56.9 AR (Average Recall) in 2019, to 69.8 AR in 2020, and to new heights of 83.7 AR in 2022. Out of 49 pose estimation methods evaluated since 2019, the top 18 are from 2022. More details can be found in the BOP challenge 2022 paper.
Besides the three tasks from 2022 (object detection, segmentation and pose estimation of objects seen during training), the 2023 challenge introduces new tasks of detection, segmentation and pose estimation of objects unseen during training. In the new tasks, methods need to learn new objects during a short object onboarding stage (max 5 min per object) and then recognize the objects in images of diverse environments. Such methods are of a high practical relevance as they do not require expensive training for every new object, which is required by most existing methods and severely limits their scalability. This year, methods are provided 3D mesh models during the onboarding stage (next years, we are planning to introduce an even more challenging variant where only a few reference images of each object will be provided).
An implicit goal of BOP is to identify the best synthetic-to-real domain transfer techniques. The capability of methods to effectively train on synthetic images is crucial as collecting ground-truth object poses for real images is prohibitively expensive. In 2020, to foster the progress, we joined the development of BlenderProc, an open-source synthesis pipeline, and used it to generate photorealistic training images for the benchmark datasets. Methods trained on these images achieved major performance gains on real test images. However, we can still observe a performance drop due to the domain gap between the synthetic training and real test images. We therefore encourage participants to build on top of BlenderProc and publish their solutions.The challenge will be announced soon. Please stay tuned.
We invite paper submissions about unpublished work. If accepted, the papers will be published in the ICCV workshop proceedings and presented at the workshop. The papers must follow the format of the main conference (LaTeX template) and be submitted to CMT (a link will be provided).
The covered topics include but are not limited to:
Paper submission deadline: July 24, 2023 (11:59PM PST)
Paper acceptance notification: August 4, 2023
Paper camera-ready version: August 10, 2023 (11:59PM PST)
Deadline for submissions to the BOP Challenge 2023: September 26, 2023 (11:59PM UTC)
Workshop date: October 3 (PM), 2023
Tomáš Hodaň, Reality Labs at Meta, firstname.lastname@example.org
Martin Sundermeyer, Google, email@example.com
Yann Labbé, Reality Labs at Meta, firstname.lastname@example.org
Gu Wang, Tsinghua University
Eric Brachmann, Niantic
Bertram Drost, MVTec
Lingni Ma, Reality Labs at Meta
Sindi Shkodrani, Reality Labs at Meta
Ales Leonardis, University of Birmingham
Carsten Steger, Technical University of Munich, MVTec
Vincent Lepetit, ENPC ParisTech, Technical University Graz
Jiří Matas, Czech Technical University in Prague