VI-Net

Code for "VI-Net: Boosting Category-level 6D Object Pose Estimation via Learning Decoupled Rotations on the Spherical Representations", ICCV 2023. [Arxiv]

Created by Jiehong Lin, Zewei Wei, Yabin Zhang, Kui Jia.

Requirements

The code has been tested with

• python 3.7.6
• pytorch 1.9.0
• CUDA 11.3

Other dependencies:

sh dependencies.sh

Data Processing

Please refer to our another work of Self-DPDN.

Network Training

Train VI-Net for rotation estimation:

python train.py --gpus 0 --dataset ${DATASET} --mode r

Train the network of pointnet++ for translation and size estimation:

python train.py --gpus 0 --dataset ${DATASET} --mode ts 

The string "DATASET" could be set as DATASET=REAL275 or DATASET=CAMERA25.

Evaluation

To test the model, please run:

python train.py --gpus 0 --dataset ${DATASET}

The string "DATASET" could be set as DATASET=REAL275 or DATASET=CAMERA25.

Acknowledgements

Our implementation leverages the code from NOCS, DualPoseNet, and SPD.

License

Our code is released under MIT License (see LICENSE file for details).

Contact

[email protected]