XYZ-IBD: A High-precision Bin-picking Dataset for Object 6D Pose Estimation Capturing Real-world Industrial Complexity
Welcome to the official repository of XYZ-IBD, a bin-picking dataset that captures real-world industrial complexity for 6D pose estimation.
🔥 Note in this repo, we provide the Docker usage for our bin-picking simulation rendering. The code is based on the BlenderProc.
- 🔗 Project Page: xyz-ibd.github.io/XYZ-IBD
- 🤗 Hugging Face: bop-benchmark/xyzibd
- 📄 Full Paper: Arxiv
-
ICCV 2025 Worokshop - 10th International Workshop on Recovering 6D Object Pose (R6D)
-
ICCV 2025 Worokshop - TRICKY 2025: Transparent & Reflective objects In the wild Challenges
We provide reference implementations to kick-start research and benchmarking. Please refer to the baseline repository for the running details.
-
📊 BOP Toolkit Integration
XYZ-IBD is prepared with BOP format(see here for the data format details), and is fully compatible with the bop_toolkit for:- Data loading and visualization
- Evaluation on BOP Metrics
-
📊 2D Detection and 6D Pose Evaluation Server
Submit your predictions for evaluation on the official BOP server:
https://bop.felk.cvut.cz/login/?next=/sub_upload/ -
📊 Monocular Depth Evaluation
Participate in monocular depth leaderboard on Codalab:
https://codalab.lisn.upsaclay.fr/competitions/22870
cd docker
docker build -t "render_xyz_synthetic" .docker run --gpus all render_xyz_synthetic
cd "examples\datasets\deep-tote"
### Modify the OUTPUT_DIR and OBJECT FILE in config.yaml to your custom path before run scripts!
### Generate the whole synthetic dataset for all objects
blenderproc run render.py --config=config.yaml
### Generate the synthetic dataset for specific object
blenderproc run custom.py --config=config_photoneo_qiuxiao1.yaml
### Post Process the synthetic dataset to BOP Format
blenderproc run gen_masks.py --config=config_photoneo_qiuxiao1.yamlIf you find this dataset helpful, please cite this paper
@misc{huang2025xyzibdhighprecisionbinpickingdataset,
title={XYZ-IBD: High-precision Bin-picking Dataset for Object 6D Pose Estimation Capturing Real-world Industrial Complexity},
author={Junwen Huang and Jizhong Liang and Jiaqi Hu and Martin Sundermeyer and Peter KT Yu and Nassir Navab and Benjamin Busam},
year={2025},
eprint={2506.00599},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2506.00599},
}
For questions, issues, or collaboration inquiries, feel free to open an issue or reach out via the project page.

