Learning Better Keypoints for Multi-Object 6DoF Pose Estimation Yangzheng Wu, Michael Greenspan WACV 2024

Install PyTorch prior to other python packages: Install the rest dependencies by using conda:

conda env create -f environment.yml
conda activate keygnet

Download BOP core datasets from BOP website.

Download the pre-trained keypoints from 'keypoints' folder and unzip it into the root of each dataset.

python train.py --dataset dname --batch_size 32 --keypointsNo 3 --lr 1e-3 --data_root "PathToData" --ckpt_root "PathToLogs"

Test with the keyGNet keypoints by using RCVPose, PVNet, and PVN3D github repositories.

If you find our work useful in your research, please consider citing:

@inproceedings{wu2024keygnet,
  title={Learning Better Keypoints for Multi-Object 6DoF Pose Estimation},
  author={Wu, Yangzheng and Greenspan, Michael},
  booktitle={WACV},
  year={2024}
}