- This is the official implementation of the paper: Category-Level 6D Object Pose Estimation via Cascaded Relation and Recurrent Reconstruction Networks (IROS 2021).
Requires:
- Python ≥ 3.6
- Pytorch ≥ 1.0
- CUDA ≥ 9.0
Compile nn_distance:
ROOT=/path/to/object-deformnet
cd $ROOT/lib/nn_distance
python setup.py install --user
Please download the Dataset following SPD.
Unzip and organize these files in $ROOT as follows:
data
├── CAMERA
│ ├── train
│ └── val
├── Real
│ ├── train
│ └── test
├── gts
│ ├── val
│ └── real_test
└── obj_models
├── train
├── val
├── real_train
└── real_test
results
├── mrcnn_results
│ ├── real_test
│ └── val
└── nocs_results
├── real_test
└── test
# train: python train.py
# test: python evaluate.py
Please download our models $ROOT/models.
Camera25:
| Camera25 | 3D50 | 3D75 | 5°2cm | 5°5cm | 10°2cm | 10°5cm |
|---|---|---|---|---|---|---|
| Our Paper | 93.8 | 88.0 | 72.0 | 76.4 | 81.0 | 87.7 |
| Released Model | 93.7 | 88.1 | 71.2 | 76.5 | 81.1 | 88.4 |
Real275:
| Real275 | 3D50 | 3D75 | 5°2cm | 5°5cm | 10°2cm | 10°5cm |
|---|---|---|---|---|---|---|
| Our Paper | 79.3 | 55.9 | 27.8 | 34.3 | 47.2 | 60.8 |
| Released Model | 78.9 | 57.8 | 31.5 | 34.9 | 51.5 | 62.8 |
If you find our work helpful, please consider citing:
@inproceedings{wang2021category,
title={Category-level 6d object pose estimation via cascaded relation and recurrent reconstruction networks},
author={Wang, Jiaze and Chen, Kai and Dou, Qi},
booktitle={2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
pages={4807--4814},
year={2021},
organization={IEEE}
}
You can also refer to our last work SGPA on ICCV2021：
@inproceedings{chen2021sgpa,
title={Sgpa: Structure-guided prior adaptation for category-level 6d object pose estimation},
author={Chen, Kai and Dou, Qi},
booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
pages={2773--2782},
year={2021}
}
WANG Jiaze: jzwang.cuhk@gmail.com | Homepage
Our implementation leverages the code from SPD, NOCS, 3PU and DCP.