Skip to content
forked from yufu-wang/6D_Pose

Implement 6D object pose estimation from keypoints

License

Notifications You must be signed in to change notification settings

Divyam10/6D_Pose

 
 

Repository files navigation

6D_Pose

Python implementation for the BOP benchmark section of the paper:
Semantic keypoint-based pose estimation from single RGB frames
Field Robotics
[Paper] cover

Data

You can download the pretrained models for detection and keypoint detection. Please place the models as follows. We also put the test images for the LMO dataset in this repo for convenience.

- data
-- detect_checkpoints
-- kpts_checkpoints

Demo

Our method uses additional 3D keypoint annotation on the CAD models, which is included in kpts_3d.json. We provide two demo. To explore the 3D annotation, please use demo_data.ipynb. To explore the inference pipeline, please use demo_pipeline.ipynb.

Reference

@article{schmeckpeper2022semantic,
  Title          = {Semantic keypoint-based pose estimation from single RGB frames},
  Author         = {Schmeckpeper, Karl and Osteen, Philip R and Wang, Yufu and Pavlakos, Georgios and Chaney, Kenneth and Jordan, Wyatt and Zhou, Xiaowei and Derpanis, Konstantinos G and Daniilidis, Kostas},
  Booktitle      = {Field Robotics},
  Year           = {2022}
}

About

Implement 6D object pose estimation from keypoints

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 73.8%
  • Python 26.2%