ROBI_Toolkit

Python scripts to work on the ROBI dataset - a multi-view dataset for reflective objects in robotic bin-picking.
ROBI dataset is available at: https://www.trailab.utias.utoronto.ca/robi.

Requirements

Python >=3.5
opencv-python >= 3.1
numpy
ruamel.yaml

Baseline Methods

We provide the evaluation results on three object pose estimators (PPF [1], Line2D [2], AAE [3]), reasearchers are welcome to compare them against their our methods. The raw evaluation results can be downloaded here. Please run "eval_baselines.py" to load the the estimated object poses from these baseline approaches.

Code

• "read_scene_imgs.py": A script to load test images (with 6D camera poses and the ground truth 6D object poses).
• "read_pattern_imgs.py": A script to load stereo pattern images and disparity maps.
• "eval_baselines.py": A script to load ground truth and the estimated object poses from provided baseline methods.

Ground Truth Depth Map

Please note that, we capture the ground truth depth map with only Ensenso camera (no Realsense data).

• For Scene 4, 5, 8, 9, each viewpoint image has the corresponding GT depth map (in "GT_Depth" folder).
• For Scene 1, 2, 3, 6, 7, the GT depth maps were captured only for a subset of viewpoints in Ensenso data folder:
  • Scene 1-3: DEPTH_view_{71-87}
  • Scene 6-7: DEPTH_view_{12-14, 16, 18, 22, 24-41}

Author

Jun Yang
[email protected]
Institute for Aerospace Studies, University of Toronto

References

[1] Drost, Bertram, et al. "Model globally, match locally: Efficient and robust 3D object recognition." 2010 IEEE computer society conference on computer vision and pattern recognition. Ieee, 2010.
[2] Hinterstoisser, Stefan, et al. "Model based training, detection and pose estimation of texture-less 3d objects in heavily cluttered scenes." Asian conference on computer vision. Springer, Berlin, Heidelberg, 2012.
[3] Sundermeyer, Martin, et al. "Implicit 3d orientation learning for 6d object detection from rgb images." Proceedings of the european conference on computer vision (ECCV). 2018.

Citation

If you find ROBI dataset useful in your work, please consider citing:

@inproceedings{yang2021robi,
  title={ROBI: A Multi-View Dataset for Reflective Objects in Robotic Bin-Picking},
  author={Yang, Jun and Gao, Yizhou and Li, Dong and Waslander, Steven L},
  booktitle={2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  pages={9788--9795},
  year={2021},
  organization={IEEE}
}

@article{yang2021probabilistic,
  title={Probabilistic Multi-View Fusion of Active Stereo Depth Maps for Robotic Bin-Picking},
  author={Yang, Jun and Li, Dong and Waslander, Steven L},
  journal={IEEE Robotics and Automation Letters},
  volume={6},
  number={3},
  pages={4472--4479},
  year={2021},
  publisher={IEEE}
}