11th Workshop on
Recovering 6D Object Pose (R6D)
Organized at ECCV 2026, September 8, 8:00-12:00 CEST, Malmö, Sweden
Organized at ECCV 2026, September 8, 8:00-12:00 CEST, Malmö, Sweden
The 2026 edition of the R6D workshop primarily focuses on visual grounding capabilities of recent vision-language models (VLMs). BOP Challenge 2026, organized together with the workshop, brings this theme into a concrete evaluation setting through the new BOP-Refer benchmark, where the goal is to localize objects (in 3D or 2D) given an input image with known intrinsics and a natural-language referring expression.
This theme extends the long-running R6D agenda covering topics from object-centric computer vision such as 6DoF object pose estimation and tracking, 3D object modeling and reconstruction, synthesis of effective training data, hand-object interaction, or robotic grasping.
Previous workshop editions: 1st edition (ICCV'15), 2nd edition (ECCV'16), 3rd edition (ICCV'17), 4th edition (ECCV'18), 5th edition (ICCV'19), 6th edition (ECCV'20), 7th edition (ECCV'22), 8th edition (ICCV'23), 9th edition (ECCV'24), 10th edition (ICCV'25).
This year the BOP challenge focuses on monocular 3D (and 2D) object grounding. Given an image with known camera intrinsics and a text query referring to one or more visible objects, the task is to predict 3D (or 2D) bounding boxes of the referred objects.
Methods are evaluated on the new BOP-Refer benchmark (to be released soon), which also comes with a first systematic evaluation of frontier vision-language models, revealing two very different regimes. 2D grounding is solid but far from saturated, with the best model reaching 53.7% AP. On the other hand, monocular 3D grounding remains largely unsolved, with the best model, Qwen3-VL, reaching only 2.6% AP on the same images and queries, leaving large headroom for the challenge. The examples below show Qwen3-VL predictions (red), one of the top-performing methods, against the ground truth (green).
“the cordless Ryobi tools”
“the light blue toy car with a white roof”
“the two tomato sauce cans”
“T shaped bracket closest to bottom left corner”
“the largest object on the table”
“two matching triangular plates on the left”
“decorative figurines that are not tools”
“the items behind the soup can”
BOP 2017–2025 summary: We created the BOP benchmark in 2017 with the goal of measuring progress in 6DoF object pose estimation and related tasks, and have been organizing challenges on the benchmark datasets ever since. We have witnessed major improvements in model-based object pose estimation: the accuracy has increased by more than 50% since 2019, competitive results have been achieved when training only on synthetic images, and the top methods for unseen objects (CAD models not available at training) have started to reach the accuracy of methods for seen objects. In 2024, we introduced model-free tasks and the BOP-H3 datasets, where objects are onboarded from reference images instead of CAD models. In 2025, we added the BOP-Industrial datasets for industrial robotics, where multi-view clearly outperforms single-view.
We invite paper submissions about unpublished work. If accepted, the papers will be published in the ECCV workshop proceedings and presented at the workshop.
The papers must have 7–14 pages, follow the submission policies of the main conference (with the exception of the number of pages), and be submitted to the OpenReview system (TBA).
The covered topics include but are not limited to:
| 8:00 | Opening |
| TBA | Invited talk 1: TBA |
| TBA | Invited talk 2: TBA |
| TBA | Coffee break and workshop posters |
| TBA | Invited talk 3: TBA |
| TBA | Invited talk 4: TBA |
| TBA | BOP Challenge 2026 results |
| TBA | Workshop paper presentations and poster session |
| 12:00 | End of workshop |
Workshop date: September 8, 2026, 8:00-12:00 CEST
Workshop paper submission deadline: July 24, 2026
Workshop paper acceptance notification: August 7, 2026
Workshop paper camera-ready version: August 14, 2026
BOP Challenge 2026 opening: TBA
BOP Challenge 2026 final submission deadline: November 27, 2026
